{
  "source_file": "0930",
  "source_platform": "Substack",
  "author": "swyx & Alessio Fanelli / Latent Space",
  "title_listed": "The 2025 AI Engineer Reading List",
  "keywords": "AI工程,学习路线,技术栈",
  "domain": "大模型算法",
  "content_type": "资源列表",
  "is_paid_flag": "否",
  "url": "https://www.latent.space/p/2025-papers",
  "publish_date": "2025-01",
  "corpus_stage": "SFT",
  "notes": "AI工程师年度推荐阅读，覆盖全技术栈",
  "id": "fd41a5f828b9",
  "collected_at": "2026-10-01T19:49:51",
  "title": "The 2025 AI Engineering Reading List - Latent.Space",
  "body_markdown": "Discussions on\nX\n,\nLinkedIn\n,\nYouTube\n. Also: Meet AI Engineers in person! Applications closing soon for attending and sponsoring\nAI Engineer Summit NYC\n, Feb 20-21.\nThe picks from all the speakers in our\nBest of 2024 series\ncatches you up for 2024, but since we wrote about running\nPaper Clubs\n, we’ve been asked many times for a reading list to recommend for those starting from scratch at work or with friends. We started with\nthe 2023 a16z Canon\n, but it needs a 2025 update and a practical focus.\nHere we curate “required reads” for the AI engineer. Our design goals are:\npick ~50 papers\n(~one\n1\na week for a year), optional extras. Arbitrary constraint.\ntell you\nwhy\nthis paper matters\ninstead of just name drop without helpful context\nbe\nvery\npractical\nfor the AI Engineer; no time wasted on\nAttention is All You Need\n, bc 1)\neveryone\nelse\nalready starts there, 2) most won’t really need it at work\nLatent Space is a\nreader-supported publication for AI Engineers\n! We really appreciate you sharing and supporting our work.\nSubscribe\nWe ended up picking 5 “papers” per section for:\nSection 1: Frontier LLMs\nSection 2: Benchmarks and Evals\nSection 3: Prompting, ICL & Chain of Thought\nSection 4: Retrieval Augmented Generation\nSection 5: Agents\nSection 6: Code Generation\nSection 7: Vision\nSection 8: Voice\nSection 9: Image/Video Diffusion\nSection 10: Finetuning\nSection 1: Frontier LLMs\nGPT1\n,\nGPT2\n,\nGPT3\n,\nCodex\n,\nInstructGPT\n,\nGPT4\npapers\n. Self explanatory.\nGPT3.5\n,\n4o\n,\no1\n,\no3\n,\nGPT4.5\nhave launch events and system cards\n2\ninstead.\nClaude 3\n/\n4\nand\nGemini 1\n/\n2.5\nto understand the leading labs. You should also know\nClaude 3.5 Sonnet\nand\nGemini 2.0 Flash\n/\nFlash Thinking\n. Also\nGemma 2\n&\n3\n(you can read the\nPyTorch re-implementation\n).\nLLaMA 1\n,\nLlama 2\n,\nLlama 3\npapers\nto understand the leading open models. You can also view\nMistral 7B\n,\nMixtral\nand\nPixtral\nas a branch on the Llama family tree. More recently China models have overtaken:\nKimi K-2\nand\nQwen 3\n.\nDeepSeek\nV1\n,\nCoder\n,\nMath\n(esp\nGRPO\n)\n,\nMoE\n,\nV2\n,\nV3\n,\nR1\npapers\n.\nPost Training Survey Papers\nlike\nthis MBZUAI one\n,\nthis one\n, and others.\nYou can both use and learn a lot from other LLMs, this is a vast topic. Bonus picks:\nIn particular, BERTs are underrated as workhorse classification models - see\nModernBERT\n/\nNeoBERT\nfor the state of the art, and\nColBERT for applications\n.\n(1.5hr video)\nCompare architectures\nof the various open models from\nAhead of AI\n.\nHonorable mentions of LLMs to know: AI2 (\nOlmo\n,\nMolmo\n,\nOlmOE\n,\nTülu\n3,\nOlmo 2\n),\nGrok\n(esp\nGrok 3\n),\nAmazon Nova\n,\nYi\n,\nReka\n,\nCohere\n,\nNemotron\n,\nMicrosoft Phi\n,\nHuggingFace SmolLM\n(esp\nSmolLM 3\n),\nApple Intelligence\n- mostly lower in ranking or lack papers.\nResearch to know: If time allows, we recommend the\nScaling Laws\nliterature:\nKaplan\n,\nChinchilla\n,\nEmergence\n/\nMirage\n,\nPost-Chinchilla laws\n.\n3\nAs of July 2025,\nKimi K-2/Moonshot\nand\nMuon\nis also strongly recommended.\nIn 2025, the frontier (o1,\no3\n,\nR1\n,\nQwQ\n/\nQVQ\n,\nf1\n) will be very much dominated by\nreasoning models\n, where\nSebastian Raschka\ncurrently has the best literature review. The basic knowledge is\nLet’s Verify Step By Step\n4\n,\ns1\n,\nSTaR\n(\ncore to xAI\n), and\nNoam Brown’s talks/podcasts\n. Most practical knowledge is\naccumulated by outsiders\n(\nLS talk\n) and tweets.\nSection 2: Benchmarks and Evals\nMMLU\npaper\n- the main\nknowledge\nbenchmark, next to\nGPQA\nand\nBIG-Bench\n. In 2025 frontier labs use\nMMLU Pro\n,\nGPQA Diamond\n, and\nBIG-Bench Hard\n.\nMRCR\n5\npaper\n- evaluating\nlong context\n,\nused by OpenAI\nover\nMuSR\n,\nLongBench\n,\nBABILong\n, and\nRULER\n. Fixing the overused\nNeedle in a Haystack\n.\nMATH\npaper\n- a compilation of\nmath competition problems\n. Frontier labs focus on\nFrontierMath\n,\nAMO\nand subsets of MATH: MATH level 5,\nAIME\n,\nAMC10/AMC12\n.\nIFEval\npaper\n- the leading\ninstruction following\neval. See also\nFacebook’s\nMulti-IF\n,\nCOLLIE\nand\nScale MultiChallenge\n, which has now overtaken\nMT-Bench\n.\nARC AGI\nchallenge\n- a famous\nabstract reasoning\n“IQ test” benchmark that has lasted far longer than many quickly saturated benchmarks.\nThis is a moving and\noften gamed\ntarget -\nMay 2025 update\nhere. We covered many of these in\nBenchmarks 101\nand\nBenchmarks 201\n, while our\nCarlini\n,\nLMArena\n(\nwarning\n!), and\nBraintrust\ncovered private, arena, and product evals (read\nLLM-as-Judge\n- and\nRubrics\n-and\nApplied LLMs essay\n). Benchmarks are linked to\nDatasets\n.\nThe bigger picture is that benchmarks saturate increasingly quickly, and there is a sense that\nthe whole approach is getting dated\nand\nnew approaches are needed\n.\nSection 3: Prompting, ICL & Chain of Thought\nNote\n: The GPT3 paper (“Language Models are Few-Shot Learners”) should already have introduced In-Context Learning (ICL) - a close cousin of prompting. We also consider\nprompt injections\nrequired knowledge —\nLilian Weng\n,\nSimon W\n.\nThe Prompt Report\npaper\n- a survey of prompting papers (\npodcast\n).\nChain-of-Thought\npaper\n- one of multiple claimants to popularizing\nChain of Thought\n, along with\nScratchpads\nand\nLet’s Think Step By Step\nTree of Thought\npaper\n-\nintroducing\nlookaheads\nand\nbacktracking\n(\npodcast\n)\nPrompt Tuning\npaper -\nyou may not need prompts - if you can do\nPrefix-Tuning\n,\nadjust decoding\n(say\nvia entropy\n), or\nrepresentation engineering\nAutomatic Prompt Engineering\npaper\n- it is increasingly obvious that\nhumans are terrible zero-shot prompters\nand\nprompting itself\ncan be enhanced by LLMs. The most notable implementation of this is in the\nDSPy paper\n/framework.\nMore survey papers on\nreasoning\n(\nslides\n). Prompting and\nContext Engineering\nis one area where reading disparate papers may not be as useful as having more practical guides - we recommend\nLilian Weng\n,\nEugene Yan\n, and\nAnthropic’s Prompt Engineering Tutorial\nand\nAI Engineer Workshop\n.\nSection 4: Retrieval Augmented Generation\nIntroduction to Information Retrieval\n- a bit unfair to recommend a book, but we are trying to make the point that RAG is an IR problem and\nIR has a 60 year history\nthat includes\nTF-IDF\n,\nBM25\n,\nFAISS\n,\nHNSW\nand other “boring” techniques.\n2020\nMeta RAG\npaper\n- which coined the term. The original authors have started Contextual and have\ncoined RAG 2.0\n. Modern “table stakes” for RAG —\nHyDE\n,\nchunking\n,\nrerankers\n,\nmultimodal data\nare\nbetter\npresented\nelsewhere\n.\nMTEB\npaper\n-\nknown overfitting\nthat\nits author considers it dead\n, but still de-facto benchmark. Many embeddings have papers - pick your poison -\nSentenceTransformers\n,\nOpenAI\n,\nNomic Embed, Jina v3, cde-small-v1\n,\nModernBERT Embed\n- with\nMatryoshka embeddings\nincreasingly standard.\nGraphRAG\npaper\n-\nMicrosoft’s\ntake on adding knowledge graphs to RAG,\nnow open sourced\n. One of the\nmost popular trends in RAG\nin 2024, alongside of\nColBERT\n/ColPali/ColQwen (more in the Vision section).\nRAGAS\npaper -\nthe simple RAG eval\nrecommended by OpenAI\n. See also\nNvidia FACTS framework\nand\nExtrinsic Hallucinations in LLMs\n- Lilian Weng’s survey of causes/evals for hallucinations (see also\nJason Wei on recall vs precision\n).\nRAG is the bread and butter of AI Engineering at work in 2024, so there are a LOT of industry resources and practical experience you will be expected to have.\nLlamaIndex\n(\ncourse\n) and\nLangChain\n(\nvideo\n) have perhaps invested the most in educational resources. You should also be familiar with the perennial\nRAG vs Long Context\ndebate.\nSection 5: Agents\nSWE-Bench\n(\nour pod\n) &\nSWE-Lancer\n- after\nadoption by Anthropic\n, Devin and\nOpenAI\n, probably the highest profile agent benchmark\n6\ntoday (vs\nWebArena\nor\nSWE-Gym\n). Technically a coding benchmark, but more a test of agents than raw LLMs. See also\nSWE-Agent\n,\nSWE-Bench Multimodal\nand the\nKonwinski Prize\n. For Tool-Agent-User interaction there is\nTauBench\nfor Airlines and Retail and\nGAIA\n.\nReAct\npaper\n(\nour podcast\n) - ReAct started a long line of research on\ntool using and function calling LLMs,\nincluding\nGorilla\nand the\nBFCL Leaderboard\n. Of historical interest -\nToolformer\nand\nHuggingGPT\n.\nMemGPT\npaper\n- one of\nmany notable approaches\nto emulating long running agent memory, adopted by\nChatGPT\nand\nLangGraph\n. Versions of these are reinvented in every agent system from\nMetaGPT\nto\nAutoGen\nto\nSmallville\n.\nVoyager\npaper\n- Nvidia’s take on 3\ncognitive architecture\ncomponents (\ncurriculum, skill library, sandbox\n) to improve performance. More abstractly, skill library/curriculum can be abstracted as a form of\nAgent Workflow Memory\n.\nAnthropic on\nBuilding Effective Agents\n(\ntalk version\n) - just a great state-of-2024 recap that focuses on the importance of\nchaining\n,\nrouting, parallelization, orchestration, evaluation, and optimization.\nSee also\nLilian Weng’s Agents\n(ex OpenAI),\nShunyu Yao on LLM Agents\n(now at OpenAI) and\nChip Huyen’s Agents\n.\nWe covered many of\nthe 2024 SOTA agent designs at NeurIPS\n, and you can find more readings in\nthe UC Berkeley LLM Agents MOOC\n. Note that we skipped bikeshedding agent definitions, but if you really need one, you could\nuse mine\n.\nSection 6: Code Generation\nThe Stack paper\n- the original open dataset twin of The Pile focused on code, starting a great lineage of open codegen work from\nThe Stack v2\nto\nStarCoder\n.\nOpen Code Model papers\n- choose from\nDeepSeek-Coder\n,\nQwen2.5-Coder\n, or\nCodeLlama\n. Many regard\n3.5 Sonnet as the best code model\nbut it has no paper.\nHumanEval/Codex paper\n- This is a saturated benchmark, but is required knowledge for the code domain. SWE-Bench is more famous for coding now, but is expensive/evals agents rather than models. Modern replacements include\nAider\n,\nCodeforces\n,\nIOI\n,\nBigCodeBench\n,\nLiveCodeBench\nand\nSciCode\n.\nAlphaCodeium paper\n- Google published\nAlphaCode\nand\nAlphaCode2\nwhich did very well on programming problems, but here is one way\nFlow Engineering\ncan add a lot more performance to any given base model.\nCriticGPT\npaper\n- LLMs are\nknown\nto generate code that can have security issues. OpenAI trained CriticGPT to spot them, and Anthropic\nuses SAEs to identify LLM features\nthat cause this, but it is a problem you should be aware of.\nCodeGen is another field where much of the frontier has moved from research to industry and\npractical engineering advice on codegen\nand\ncode agents like Devin\nare only found in industry blogposts and talks rather than research papers.\nSection 7: Vision\nNon-LLM Vision work\nis still important: e.g. the\nYOLO\npaper\n(now\nup to v11\n, but\nmind the lineage\n), but increasingly transformers like\nDETRs Beat YOLOs\ntoo.\nCLIP\npaper\n- the first successful\nViT\nfrom Alec Radford. These days, superceded by\nBLIP\n/\nBLIP2\nor\nSigLIP/PaliGemma\n, but still required to know.\nMMVP benchmark\n(\nLS Live\n)- quantifies issues with CLIP. Multimodal versions of MMLU (\nMMMU\n), and\nSWE-Bench\ndo exist. See also\nMathVista\nand\nCharXiv\n.\nSegment Anything Model\nand\nSAM 2\npaper\n(\nour pod\n) - the very successful image and video segmentation foundation model. Pair with\nGroundingDINO\n.\nEarly fusion research\n: Contra the cheap “late fusion” work like\nLLaVA\n(\nour pod\n), early fusion covers Meta’s\nFlamingo\n,\nChameleon\n, Apple’s\nAIMv2\n, Reka\nCore\n, et al. In reality there\nare at least 4 streams of visual LM work\n.\nMuch frontier VLM work these days is no longer published (the last we really got was\nGPT4V system card\nand\nderivative papers\n). We recommend having working experience with vision capabilities of 4o (including\nfinetuning 4o vision\n), Claude 3.5 Sonnet/Haiku, Gemini 2.0 Flash, and o1. Others:\nPixtral\n,\nLlama 3.2\n,\nMoondream\n,\nQVQ\n.\nOCR is an important subset workhorse functionality of vision, so you may wish to check out\nMistral OCR\nand\nVLM Run\n.\nSection 8: Voice\nWhisper\npaper\n- the successful\nASR\nmodel from Alec Radford. Whisper\nv2\n,\nv3\nand\ndistil-whisper\nand\nv3 Turbo\nare open weights but have no paper.\nAudioPaLM\npaper\n- our last look at Google’s voice thoughts before PaLM became Gemini. See also: Meta’s\nLlama 3 explorations into speech\n.\nNaturalSpeech\npaper\n- one of a few leading\nTTS\napproaches. Recently\nv3\n.\nKyutai Moshi\npaper\n- an impressive full-duplex speech-text open weights model with\nhigh profile demo\n. See also\nHume OCTAVE\n.\nOpenAI Realtime API: The Missing Manual\n- Again, frontier omnimodel work is not published, but we did our best to document the Realtime API.\nWe do recommend diversifying from the big labs here for now - try Daily, Livekit, Vapi, Assembly, Deepgram, Fireworks, Cartesia, Elevenlabs etc. See\nthe State of Voice 2024\n. While NotebookLM’s voice model is not public,\nwe got the deepest description of the modeling process\nthat we know of. Historical interest:\nwav2vec\n(see Snipd pod)\nWith Gemini 2.0 also being natively voice and vision multimodal, the Voice and Vision modalities are on a clear path to merging in 2025 and beyond.\nSection 9: Image/Video Diffusion\nLatent Diffusion\npaper\n- effectively the Stable Diffusion paper. See also\nSD2\n,\nSDXL\n,\nSD3\npapers. These days the team is working on\nBFL Flux\n[schnell|dev|pro].\nDALL-E\n/\nDALL-E-2\n/\nDALL-E-3\npaper\n- OpenAI’s image generation.\nImagen\n/\nImagen 2\n/\nImagen 3\npaper\n- Google’s image gen.\nConsistency Models\npaper\n- this distillation work with\nLCMs\nspawned the\nquick draw viral moment of Dec 2023\n. These days, updated with\nsCMs\nor\nDMDs\n.\nSora\nblogpost\n- text to video - no paper of course beyond\nthe DiT paper\n(same authors), but still the most significant launch of the year, with many\nopen weights competitors\nlike\nOpenSora\n.\nLilian Weng survey here\n.\nAutoregressive image generation\nis all the rage this year for\nGemini\n,\n4o\n, and\nLlama’s\nNative Image Gen.\nMore Image/Video work is done in Startups now. See\nIdeogram\n,\nRecraft\n,\nReve Image\n, Pika, Playground,\nGenmo\nand Chinese models like Wan 2.1 and Kling.\nWe also highly recommend familiarity with\nComfyUI\n(\nwe were first to interview\n).\nText Diffusion\n(\nMercury/Inception\nis SOTA) and\nMusic Diffusion\nare niche for now.\nFor historical interest:\nDALL-E\n/\nDALL-E-2\n/\nDALL-E-3\npaper\n- OpenAI’s image generation, and\nImagen\n/\nImagen 2\n/\nImagen 3\npaper\n- Google’s image gen.\nSection 10: Finetuning\nLoRA\n/\nQLoRA\npaper\n- the de facto way to finetune models cheaply, whether on local models or with 4o (\nconfirmed on pod\n).\nFSDP+QLoRA\nis educational.\nDPO\npaper\n- the popular, if slightly inferior, alternative to\nPPO\n, now supported by OpenAI as\nPreference Finetuning\n.\nReFT\npaper\n- instead of finetuning a few layers, focus on features instead.\nOrca 3/AgentInstruct\npaper\n- see the\nSynthetic Data picks at NeurIPS\nbut this is a great way to get finetue data.\nRL/Reasoning Tuning papers\n-\nRL Finetuning for o1\nis debated, but\nLet’s Verify Step By Step\nand\nNoam Brown’s many public talks\ngive hints for how it works.\nWe recommend going thru\nthe Unsloth notebooks\nand HuggingFace’s\nHow to fine-tune open LLMs\nfor more on the full process. This is obviously an endlessly deep rabbit hole that, at the extreme, overlaps with the Research Scientist track.\nHow To Start",
  "metadata": {
    "title": "The 2025 AI Engineering Reading List - Latent.Space",
    "description": "We picked 50 paper/models/blogs across 10 fields in AI Eng: LLMs, Benchmarks, Prompting, RAG, Agents, CodeGen, Vision, Voice, Diffusion, Finetuning. If you're starting from scratch, start here.",
    "keywords": null,
    "author": "Latent.Space",
    "og:type": "article",
    "og:title": "The 2025 AI Engineering Reading List",
    "og:description": "We picked 50 paper/models/blogs across 10 fields in AI Eng: LLMs, Benchmarks, Prompting, RAG, Agents, CodeGen, Vision, Voice, Diffusion, Finetuning. If you're starting from scratch, start here.",
    "og:image": "https://substackcdn.com/image/fetch/$s_!PQFf!,w_1200,h_675,c_fill,f_jpg,q_auto:good,fl_progressive:steep,g_auto/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9242370c-229f-453d-924a-7a5aa4d20a4c_1090x502.png",
    "og:url": "https://www.latent.space/p/2025-papers",
    "twitter:title": "The 2025 AI Engineering Reading List",
    "twitter:description": "We picked 50 paper/models/blogs across 10 fields in AI Eng: LLMs, Benchmarks, Prompting, RAG, Agents, CodeGen, Vision, Voice, Diffusion, Finetuning. If you're starting from scratch, start here.",
    "twitter:image": "https://substackcdn.com/image/fetch/$s_!SN75!,f_auto,q_auto:best,fl_progressive:steep/https%3A%2F%2Fswyx.substack.com%2Fapi%2Fv1%2Fpost_preview%2F153615976%2Ftwitter.jpg%3Fversion%3D4",
    "twitter:card": "summary_large_image",
    "article:modified_time": "2025-12-29T07:45:07.273Z"
  },
  "paywall": {
    "listed_paid": false,
    "detected": false,
    "marker": "",
    "public_only": false,
    "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙"
  },
  "crawl": {
    "success": true,
    "status_code": 200,
    "error": ""
  },
  "code_blocks": [],
  "images": [
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!dhbm!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1951ae5c-bb0a-4527-bba8-f234e65f6cc4_1812x822.jpeg",
      "bytes": 79878,
      "reused": true,
      "local_path": "attachments/fd41a5f828b9/images/014.jpeg",
      "src": "https://substackcdn.com/image/fetch/$s_!dhbm!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1951ae5c-bb0a-4527-bba8-f234e65f6cc4_1812x822.jpeg",
      "alt": "progress",
      "downloaded": true
    }
  ],
  "audio": [],
  "videos": [
    {
      "url": "https://www.youtube.com/watch?v=hnIMY9pLPdg",
      "status": "ok",
      "local_path": "attachments/fd41a5f828b9/videos/hnIMY9pLPdg.mp4",
      "bytes": 460676566,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "3726.021000"
    },
    {
      "url": "https://www.youtube.com/watch?v=wT636THdZZo",
      "src": "https://www.youtube.com/watch?v=wT636THdZZo&list=PLWEAb1SXhjlfG63F03R52DZXpHzVB1_5j",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-wT636THdZZo.mp4",
      "bytes": 1755700011,
      "valid": true,
      "height": 720,
      "width": 1280,
      "duration": "31884.041000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=zjkBMFhNj_g",
      "src": "https://www.youtube.com/watch?v=zjkBMFhNj_g",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-zjkBMFhNj_g.mp4",
      "bytes": 120860496,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "3587.701000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=XCgWoLu-zK8",
      "src": "https://youtu.be/XCgWoLu-zK8",
      "status": "no_1080_or_720",
      "kind": "single_video",
      "error": "ERROR: [youtube] XCgWoLu-zK8: Requested format is not available. Use --list-formats for a list of available formats\n",
      "title": "",
      "retried_at": "2026-10-20T17:48:31Z"
    },
    {
      "url": "https://www.youtube.com/watch?v=Y5-FeaFOEFM",
      "src": "https://www.youtube.com/watch?v=Y5-FeaFOEFM",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-Y5-FeaFOEFM.mp4",
      "bytes": 266377486,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "2831.161000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=Gr_eYXdHFis",
      "src": "https://www.youtube.com/live/Gr_eYXdHFis",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-Gr_eYXdHFis.mp4",
      "bytes": 317308336,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "3122.941000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=skT89EvIjrc",
      "src": "https://www.youtube.com/watch?v=skT89EvIjrc&t=68s",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-skT89EvIjrc.mp4",
      "bytes": 38750669,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "981.781000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=hkhDdcM5V94",
      "src": "https://www.youtube.com/watch?v=hkhDdcM5V94",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-hkhDdcM5V94.mp4",
      "bytes": 686920981,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "5695.581000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=i2vBaFzCEJw",
      "src": "https://www.youtube.com/watch?v=i2vBaFzCEJw",
      "status": "ok",
      "kind": "single_video",
      "title": "",
      "local_path": "attachments/fd41a5f828b9/videos/repair-i2vBaFzCEJw.mp4",
      "bytes": 123238388,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "991.701000",
      "reused": true,
      "retried_at": "2026-10-18T17:15:18Z"
    },
    {
      "url": "https://www.youtube.com/watch?v=TRjq7t2Ms5I",
      "src": "https://www.youtube.com/watch?v=TRjq7t2Ms5I&t=152s",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-TRjq7t2Ms5I.mp4",
      "bytes": 44401918,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1114.201000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=FDEmbYPgG-s",
      "src": "https://www.youtube.com/watch?v=FDEmbYPgG-s",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-FDEmbYPgG-s.mp4",
      "bytes": 86661004,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "739.061000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=DId2KP8Ykz4",
      "src": "https://www.youtube.com/watch?v=DId2KP8Ykz4",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-DId2KP8Ykz4.mp4",
      "bytes": 75613825,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1093.061000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=VIqXNRsRRQo",
      "src": "https://www.youtube.com/watch?v=VIqXNRsRRQo",
      "status": "no_1080_or_720",
      "kind": "single_video",
      "error": "ERROR: [youtube] VIqXNRsRRQo: Requested format is not available. Use --list-formats for a list of available formats\n",
      "title": "",
      "retried_at": "2026-10-20T17:48:31Z"
    },
    {
      "url": "https://www.youtube.com/watch?v=knDDGYHnnSI",
      "src": "https://www.youtube.com/watch?v=knDDGYHnnSI",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-knDDGYHnnSI.mp4",
      "bytes": 57709979,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1154.381000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=wd7TZ4w1mSw",
      "src": "https://www.youtube.com/watch?v=wd7TZ4w1mSw",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-wd7TZ4w1mSw.mp4",
      "bytes": 9161949,
      "valid": true,
      "height": 720,
      "width": 1108,
      "duration": "312.741000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=wnsZ7DuqYp0",
      "src": "https://www.youtube.com/watch?v=wnsZ7DuqYp0",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-wnsZ7DuqYp0.mp4",
      "bytes": 155224300,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "580.221000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=Ve-akpov78Q",
      "src": "https://www.youtube.com/watch?v=Ve-akpov78Q",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-Ve-akpov78Q.mp4",
      "bytes": 112032886,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1133.421000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=T7NWjoD_OuY",
      "src": "https://www.youtube.com/watch?v=T7NWjoD_OuY&t=8s",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-T7NWjoD_OuY.mp4",
      "bytes": 114052395,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1203.841000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=T7sxvrJLJ14",
      "src": "https://www.youtube.com/watch?v=T7sxvrJLJ14",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-T7sxvrJLJ14.mp4",
      "bytes": 104106919,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "1165.061000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=hm2IJSKcYvo",
      "src": "https://www.youtube.com/watch?v=hm2IJSKcYvo",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-hm2IJSKcYvo.mp4",
      "bytes": 437044431,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "2544.281000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=yJ396Ksiv2s",
      "src": "https://www.youtube.com/watch?v=yJ396Ksiv2s",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-yJ396Ksiv2s.mp4",
      "bytes": 455539231,
      "valid": true,
      "height": 1080,
      "width": 1728,
      "duration": "2219.301000",
      "reused": false
    },
    {
      "url": "https://www.youtube.com/watch?v=1mG678f1ZYU",
      "src": "https://www.youtube.com/watch?v=1mG678f1ZYU&pp=ygUOdGV4dCBkaWZmdXNpb24%3D",
      "status": "ok",
      "kind": "single_video",
      "local_path": "attachments/fd41a5f828b9/videos/repair-1mG678f1ZYU.mp4",
      "bytes": 145474717,
      "valid": true,
      "height": 720,
      "width": 1280,
      "duration": "2543.641000",
      "reused": false
    }
  ],
  "links": [
    {
      "href": "https://x.com/latentspacepod/status/1872719928618565646",
      "label": "X",
      "type": "a"
    },
    {
      "href": "https://www.linkedin.com/posts/drjimfan_lets-start-the-new-year-with-a-buffet-of-activity-7280638043932270592-Ge2n/?utm_source=share&utm_medium=member_desktop",
      "label": "LinkedIn",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=hnIMY9pLPdg",
      "label": "YouTube",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2025-summit",
      "label": "AI Engineer Summit NYC",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=wT636THdZZo&list=PLWEAb1SXhjlfG63F03R52DZXpHzVB1_5j",
      "label": "Best of 2024 series",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/paperclub",
      "label": "Paper Clubs",
      "type": "a"
    },
    {
      "href": "https://a16z.com/ai-canon/",
      "label": "the 2023 a16z Canon",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=zjkBMFhNj_g",
      "label": "everyone",
      "type": "a"
    },
    {
      "href": "https://github.com/elicit/machine-learning-list",
      "label": "else",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/about",
      "label": "reader-supported publication for AI Engineers",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2025-papers",
      "label": "Section 1: Frontier LLMs",
      "type": "a"
    },
    {
      "href": "https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf",
      "label": "GPT1",
      "type": "a"
    },
    {
      "href": "https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf",
      "label": "GPT2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2005.14165",
      "label": "GPT3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2107.03374",
      "label": "Codex",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2203.02155",
      "label": "InstructGPT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2303.08774",
      "label": "GPT4",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/chatgpt/",
      "label": "GPT3.5",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/hello-gpt-4o/",
      "label": "4o",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/introducing-openai-o1-preview/",
      "label": "o1",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/deliberative-alignment/",
      "label": "o3",
      "type": "a"
    },
    {
      "href": "https://www.interconnects.ai/p/gpt-45-not-a-frontier-model",
      "label": "GPT4.5",
      "type": "a"
    },
    {
      "href": "https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf",
      "label": "Claude 3",
      "type": "a"
    },
    {
      "href": "https://www-cdn.anthropic.com/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf?s=09",
      "label": "4",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2312.11805",
      "label": "Gemini 1",
      "type": "a"
    },
    {
      "href": "https://storage.googleapis.com/deepmind-media/gemini/gemini_v2_5_report.pdf",
      "label": "2.5",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/claude-sonnet",
      "label": "Claude 3.5 Sonnet",
      "type": "a"
    },
    {
      "href": "https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/",
      "label": "Gemini 2.0 Flash",
      "type": "a"
    },
    {
      "href": "https://ai.google.dev/gemini-api/docs/thinking-mode",
      "label": "Flash Thinking",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2408.00118",
      "label": "Gemma 2",
      "type": "a"
    },
    {
      "href": "https://storage.googleapis.com/deepmind-media/gemma/Gemma3Report.pdf",
      "label": "3",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=44962059",
      "label": "PyTorch re-implementation",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2302.13971",
      "label": "LLaMA 1",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2307.09288",
      "label": "Llama 2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2407.21783",
      "label": "Llama 3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.06825",
      "label": "Mistral 7B",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.04088",
      "label": "Mixtral",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2410.07073",
      "label": "Pixtral",
      "type": "a"
    },
    {
      "href": "https://moonshotai.github.io/Kimi-K2/",
      "label": "Kimi K-2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2505.09388",
      "label": "Qwen 3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.02954",
      "label": "V1",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.14196",
      "label": "Coder",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2402.03300",
      "label": "Math",
      "type": "a"
    },
    {
      "href": "https://www.interconnects.ai/p/papers-im-reading-base-model-rl-grpo",
      "label": "GRPO",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.06066",
      "label": "MoE",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2405.04434",
      "label": "V2",
      "type": "a"
    },
    {
      "href": "https://github.com/deepseek-ai/DeepSeek-V3",
      "label": "V3",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
      "label": "R1",
      "type": "a"
    },
    {
      "href": "https://youtu.be/XCgWoLu-zK8",
      "label": "this MBZUAI one",
      "type": "a"
    },
    {
      "href": "https://x.com/zainhasan6/status/1899541155924046006?s=46",
      "label": "this one",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-modernbert-small-new-retrieverclassifier/",
      "label": "ModernBERT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2502.19587",
      "label": "NeoBERT",
      "type": "a"
    },
    {
      "href": "https://www.answer.ai/posts/colbert-pooling.html",
      "label": "ColBERT for applications",
      "type": "a"
    },
    {
      "href": "https://x.com/rasbt/status/1965798055141429523?s=46",
      "label": "Compare architectures",
      "type": "a"
    },
    {
      "href": "https://open.substack.com/pub/sebastianraschka",
      "label": "Ahead of AI",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2402.00838",
      "label": "Olmo",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2409.17146",
      "label": "Molmo",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2409.02060",
      "label": "OlmOE",
      "type": "a"
    },
    {
      "href": "https://allenai.org/blog/tulu-3-technical",
      "label": "Tülu",
      "type": "a"
    },
    {
      "href": "https://x.com/soldni/status/1875266934943649808?s=46",
      "label": "Olmo 2",
      "type": "a"
    },
    {
      "href": "https://github.com/xai-org/grok-1",
      "label": "Grok",
      "type": "a"
    },
    {
      "href": "https://x.ai/news/grok-3",
      "label": "Grok 3",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-olympus-has-dropped-aka-amazon-nova/",
      "label": "Amazon Nova",
      "type": "a"
    },
    {
      "href": "https://www.wired.com/story/chinese-startup-01-ai-is-winning-the-open-source-ai-race/",
      "label": "Yi",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/yitay",
      "label": "Reka",
      "type": "a"
    },
    {
      "href": "https://cohere.com/command",
      "label": "Cohere",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-to-be-named-2748/",
      "label": "Nemotron",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2412.08905",
      "label": "Microsoft Phi",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2024-open-models",
      "label": "HuggingFace SmolLM",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/blog/smollm3",
      "label": "SmolLM 3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2407.21075",
      "label": "Apple Intelligence",
      "type": "a"
    },
    {
      "href": "http://arxiv.org/abs/2001.08361",
      "label": "Kaplan",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2203.15556",
      "label": "Chinchilla",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2206.07682",
      "label": "Emergence",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2304.15004",
      "label": "Mirage",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.00448",
      "label": "Post-Chinchilla laws",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2502.16982",
      "label": "Muon",
      "type": "a"
    },
    {
      "href": "https://en.wikipedia.org/wiki/OpenAI_o3",
      "label": "o3",
      "type": "a"
    },
    {
      "href": "https://api-docs.deepseek.com/news/news1120",
      "label": "R1",
      "type": "a"
    },
    {
      "href": "https://qwenlm.github.io/blog/qwq-32b-preview/",
      "label": "QwQ",
      "type": "a"
    },
    {
      "href": "https://qwenlm.github.io/blog/qvq-72b-preview/",
      "label": "QVQ",
      "type": "a"
    },
    {
      "href": "https://fireworks.ai/blog/fireworks-compound-ai-system-f1",
      "label": "f1",
      "type": "a"
    },
    {
      "href": "https://magazine.sebastianraschka.com/p/understanding-reasoning-llms",
      "label": "Sebastian Raschka",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2305.20050",
      "label": "Let’s Verify Step By Step",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2501.19393",
      "label": "s1",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=Y5-FeaFOEFM",
      "label": "STaR",
      "type": "a"
    },
    {
      "href": "https://x.com/shaneguML/status/1944558999342661963",
      "label": "core to xAI",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/live/Gr_eYXdHFis",
      "label": "Noam Brown’s talks/podcasts",
      "type": "a"
    },
    {
      "href": "https://hn.algolia.com/?dateRange=all&page=0&prefix=false&query=o1&sort=byPopularity&type=story",
      "label": "accumulated by outsiders",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=skT89EvIjrc&t=68s",
      "label": "LS talk",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2009.03300",
      "label": "MMLU",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2311.12022",
      "label": "GPQA",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2206.04615",
      "label": "BIG-Bench",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2406.01574",
      "label": "MMLU Pro",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2210.09261",
      "label": "BIG-Bench Hard",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/html/2409.12640v2",
      "label": "MRCR",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/quasar",
      "label": "used by OpenAI",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.16049",
      "label": "MuSR",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2412.15204",
      "label": "LongBench",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2406.10149",
      "label": "BABILong",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/gradient",
      "label": "RULER",
      "type": "a"
    },
    {
      "href": "https://github.com/gkamradt/LLMTest_NeedleInAHaystack",
      "label": "Needle in a Haystack",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2103.03874",
      "label": "MATH",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2411.04872",
      "label": "FrontierMath",
      "type": "a"
    },
    {
      "href": "https://x.com/jack_w_rae/status/1924897579122491523",
      "label": "AMO",
      "type": "a"
    },
    {
      "href": "https://www.kaggle.com/datasets/hemishveeraboina/aime-problem-set-1983-2024",
      "label": "AIME",
      "type": "a"
    },
    {
      "href": "https://github.com/ryanrudes/amc",
      "label": "AMC10/AMC12",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2311.07911",
      "label": "IFEval",
      "type": "a"
    },
    {
      "href": "https://github.com/facebookresearch/Multi-IF",
      "label": "Facebook’s",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2410.15553",
      "label": "Multi-IF",
      "type": "a"
    },
    {
      "href": "https://collie-benchmark.github.io/",
      "label": "COLLIE",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2501.17399",
      "label": "Scale MultiChallenge",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/spaces/lmsys/mt-bench",
      "label": "MT-Bench",
      "type": "a"
    },
    {
      "href": "https://arcprize.org/arc",
      "label": "ARC AGI",
      "type": "a"
    },
    {
      "href": "https://x.com/kalomaze/status/1923983739299955064",
      "label": "often gamed",
      "type": "a"
    },
    {
      "href": "https://x.com/scaling01/status/1919092778648408363",
      "label": "May 2025 update",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/benchmarks-101",
      "label": "Benchmarks 101",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/benchmarks-201",
      "label": "Benchmarks 201",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/carlini",
      "label": "Carlini",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/lmarena",
      "label": "LMArena",
      "type": "a"
    },
    {
      "href": "https://x.com/karpathy/status/1917546757929722115",
      "label": "warning",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/braintrust",
      "label": "Braintrust",
      "type": "a"
    },
    {
      "href": "https://hamel.dev/blog/posts/llm-judge/",
      "label": "LLM-as-Judge",
      "type": "a"
    },
    {
      "href": "https://x.com/iscienceluvr/status/1948235609190867054?s=46",
      "label": "Rubrics",
      "type": "a"
    },
    {
      "href": "https://applied-llms.org/",
      "label": "Applied LLMs essay",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/datasets-101",
      "label": "Datasets",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!dhbm!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1951ae5c-bb0a-4527-bba8-f234e65f6cc4_1812x822.jpeg",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://ysymyth.github.io/The-Second-Half/",
      "label": "the whole approach is getting dated",
      "type": "a"
    },
    {
      "href": "https://research.trychroma.com/generative-benchmarking",
      "label": "new approaches are needed",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/i/93381455/what-is-prompt-injection",
      "label": "prompt injections",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2023-10-25-adv-attack-llm/",
      "label": "Lilian Weng",
      "type": "a"
    },
    {
      "href": "https://simonwillison.net/series/prompt-injection/",
      "label": "Simon W",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2406.06608",
      "label": "The Prompt Report",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/learn-prompting",
      "label": "podcast",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2201.11903",
      "label": "Chain-of-Thought",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2112.00114",
      "label": "Scratchpads",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2205.11916",
      "label": "Let’s Think Step By Step",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2305.10601",
      "label": "Tree of Thought",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/shunyu",
      "label": "podcast",
      "type": "a"
    },
    {
      "href": "https://aclanthology.org/2021.emnlp-main.243/",
      "label": "Prompt Tuning",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2101.00190",
      "label": "Prefix-Tuning",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2402.10200",
      "label": "adjust decoding",
      "type": "a"
    },
    {
      "href": "https://github.com/xjdr-alt/entropix",
      "label": "via entropy",
      "type": "a"
    },
    {
      "href": "https://vgel.me/posts/representation-engineering/",
      "label": "representation engineering",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2211.01910",
      "label": "Automatic Prompt Engineering",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.03714",
      "label": "DSPy paper",
      "type": "a"
    },
    {
      "href": "https://x.com/denny_zhou/status/1948499173986201915",
      "label": "reasoning",
      "type": "a"
    },
    {
      "href": "https://dennyzhou.github.io/LLM-Reasoning-Stanford-CS-25.pdf",
      "label": "slides",
      "type": "a"
    },
    {
      "href": "https://www.dbreunig.com/2025/07/24/why-the-term-context-engineering-matters.html",
      "label": "Context Engineering",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/",
      "label": "Lilian Weng",
      "type": "a"
    },
    {
      "href": "https://eugeneyan.com/writing/prompting/",
      "label": "Eugene Yan",
      "type": "a"
    },
    {
      "href": "https://github.com/anthropics/prompt-eng-interactive-tutorial",
      "label": "Anthropic’s Prompt Engineering Tutorial",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=hkhDdcM5V94",
      "label": "AI Engineer Workshop",
      "type": "a"
    },
    {
      "href": "https://nlp.stanford.edu/IR-book/information-retrieval-book.html",
      "label": "Introduction to Information Retrieval",
      "type": "a"
    },
    {
      "href": "https://en.wikipedia.org/wiki/Information_retrieval",
      "label": "IR has a 60 year history",
      "type": "a"
    },
    {
      "href": "https://en.wikipedia.org/wiki/Tf%E2%80%93idf",
      "label": "TF-IDF",
      "type": "a"
    },
    {
      "href": "https://en.wikipedia.org/wiki/Okapi_BM25",
      "label": "BM25",
      "type": "a"
    },
    {
      "href": "https://github.com/facebookresearch/faiss",
      "label": "FAISS",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/1603.09320",
      "label": "HNSW",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2005.11401",
      "label": "Meta RAG",
      "type": "a"
    },
    {
      "href": "https://contextual.ai/introducing-rag2/",
      "label": "coined RAG 2.0",
      "type": "a"
    },
    {
      "href": "https://docs.llamaindex.ai/en/stable/optimizing/advanced_retrieval/query_transformations/",
      "label": "HyDE",
      "type": "a"
    },
    {
      "href": "https://research.trychroma.com/evaluating-chunking",
      "label": "chunking",
      "type": "a"
    },
    {
      "href": "https://cohere.com/blog/rerank-3pt5",
      "label": "rerankers",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=i2vBaFzCEJw",
      "label": "multimodal data",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=TRjq7t2Ms5I&t=152s",
      "label": "better",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=FDEmbYPgG-s",
      "label": "presented",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=DId2KP8Ykz4",
      "label": "elsewhere",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2210.07316",
      "label": "MTEB",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=42504379",
      "label": "known overfitting",
      "type": "a"
    },
    {
      "href": "https://x.com/Nils_Reimers/status/1870812625505849849",
      "label": "its author considers it dead",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2",
      "label": "SentenceTransformers",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=VIqXNRsRRQo",
      "label": "OpenAI",
      "type": "a"
    },
    {
      "href": "https://x.com/zach_nussbaum/status/1873813021786767699?s=46&t=tMWvmS3OL3Ssg0b9lKvp4Q",
      "label": "ModernBERT Embed",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/blog/matryoshka",
      "label": "Matryoshka embeddings",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/pdf/2404.16130",
      "label": "GraphRAG",
      "type": "a"
    },
    {
      "href": "https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/?utm_source=ainews&utm_medium=email&utm_campaign=ainews-graphrag",
      "label": "Microsoft’s",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-graphrag/",
      "label": "now open sourced",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=knDDGYHnnSI",
      "label": "most popular trends in RAG",
      "type": "a"
    },
    {
      "href": "https://github.com/stanford-futuredata/ColBERT",
      "label": "ColBERT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2309.15217",
      "label": "RAGAS",
      "type": "a"
    },
    {
      "href": "https://x.com/swyx/status/1724490887147978793",
      "label": "recommended by OpenAI",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2407.07858v1",
      "label": "Nvidia FACTS framework",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2024-07-07-hallucination/",
      "label": "Extrinsic Hallucinations in LLMs",
      "type": "a"
    },
    {
      "href": "https://x.com/_jasonwei/status/1871285864690815053",
      "label": "Jason Wei on recall vs precision",
      "type": "a"
    },
    {
      "href": "https://docs.llamaindex.ai/en/stable/understanding/rag/",
      "label": "LlamaIndex",
      "type": "a"
    },
    {
      "href": "https://www.deeplearning.ai/short-courses/building-evaluating-advanced-rag/",
      "label": "course",
      "type": "a"
    },
    {
      "href": "https://python.langchain.com/docs/tutorials/rag/",
      "label": "LangChain",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=wd7TZ4w1mSw",
      "label": "video",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2407.16833",
      "label": "RAG vs Long Context",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.06770",
      "label": "SWE-Bench",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/iclr-2024-benchmarks-agents?utm_source=publication-search",
      "label": "our pod",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/swe-lancer/",
      "label": "SWE-Lancer",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/introducing-swe-bench-verified/",
      "label": "OpenAI",
      "type": "a"
    },
    {
      "href": "https://github.com/web-arena-x/webarena",
      "label": "WebArena",
      "type": "a"
    },
    {
      "href": "https://x.com/jiayi_pirate/status/1871249410128322856",
      "label": "SWE-Gym",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2405.15793",
      "label": "SWE-Agent",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2410.03859",
      "label": "SWE-Bench Multimodal",
      "type": "a"
    },
    {
      "href": "https://kprize.ai/",
      "label": "Konwinski Prize",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2406.12045",
      "label": "TauBench",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2311.12983",
      "label": "GAIA",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2210.03629",
      "label": "ReAct",
      "type": "a"
    },
    {
      "href": "https://gorilla.cs.berkeley.edu/",
      "label": "Gorilla",
      "type": "a"
    },
    {
      "href": "https://gorilla.cs.berkeley.edu/leaderboard.html",
      "label": "BFCL Leaderboard",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2302.04761",
      "label": "Toolformer",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2303.17580",
      "label": "HuggingGPT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.08560",
      "label": "MemGPT",
      "type": "a"
    },
    {
      "href": "https://x.com/swyx/status/1915128966203236571",
      "label": "many notable approaches",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/memory-and-new-controls-for-chatgpt/",
      "label": "ChatGPT",
      "type": "a"
    },
    {
      "href": "https://langchain-ai.github.io/langgraph/concepts/memory/",
      "label": "LangGraph",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2308.00352",
      "label": "MetaGPT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2308.08155",
      "label": "AutoGen",
      "type": "a"
    },
    {
      "href": "https://github.com/joonspk-research/generative_agents",
      "label": "Smallville",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2305.16291",
      "label": "Voyager",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2309.02427",
      "label": "cognitive architecture",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2409.07429",
      "label": "Agent Workflow Memory",
      "type": "a"
    },
    {
      "href": "https://www.anthropic.com/research/building-effective-agents",
      "label": "Building Effective Agents",
      "type": "a"
    },
    {
      "href": "https://x.com/aiDotEngineer/status/1908230651985485955",
      "label": "talk version",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2023-06-23-agent/",
      "label": "Lilian Weng’s Agents",
      "type": "a"
    },
    {
      "href": "https://huyenchip.com//2025/01/07/agents.html",
      "label": "Chip Huyen’s Agents",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2024-agents",
      "label": "the 2024 SOTA agent designs at NeurIPS",
      "type": "a"
    },
    {
      "href": "https://llmagents-learning.org/f24",
      "label": "the UC Berkeley LLM Agents MOOC",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=wnsZ7DuqYp0",
      "label": "use mine",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2211.15533",
      "label": "The Stack paper",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/datasets/bigcode/the-stack-v2",
      "label": "The Stack v2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2402.19173",
      "label": "StarCoder",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2409.12186",
      "label": "Qwen2.5-Coder",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/code-llama-open-foundation-models-for-code/",
      "label": "CodeLlama",
      "type": "a"
    },
    {
      "href": "https://aider.chat/docs/leaderboards/",
      "label": "Aider",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2312.02143",
      "label": "Codeforces",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/pdf/2502.06807v1",
      "label": "IOI",
      "type": "a"
    },
    {
      "href": "https://huggingface.co/spaces/bigcode/bigcodebench-leaderboard",
      "label": "BigCodeBench",
      "type": "a"
    },
    {
      "href": "https://livecodebench.github.io/",
      "label": "LiveCodeBench",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-to-be-named-5745/",
      "label": "SciCode",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.08500",
      "label": "AlphaCodeium paper",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=34020025",
      "label": "AlphaCode",
      "type": "a"
    },
    {
      "href": "https://x.com/RemiLeblond/status/1732419456272318614",
      "label": "AlphaCode2",
      "type": "a"
    },
    {
      "href": "https://criticgpt.org/criticgpt-openai/",
      "label": "CriticGPT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2412.15004v1",
      "label": "known",
      "type": "a"
    },
    {
      "href": "https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html",
      "label": "uses SAEs to identify LLM features",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=Ve-akpov78Q",
      "label": "practical engineering advice on codegen",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=T7NWjoD_OuY&t=8s",
      "label": "code agents like Devin",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/1506.02640",
      "label": "YOLO",
      "type": "a"
    },
    {
      "href": "https://github.com/ultralytics/ultralytics",
      "label": "up to v11",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=42352342",
      "label": "mind the lineage",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2304.08069",
      "label": "DETRs Beat YOLOs",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2103.00020",
      "label": "CLIP",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2010.11929",
      "label": "ViT",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2201.12086",
      "label": "BLIP",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2301.12597",
      "label": "BLIP2",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/i/152857207/part-vision",
      "label": "SigLIP/PaliGemma",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2401.06209",
      "label": "MMVP benchmark",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2024-vision",
      "label": "LS Live",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2311.16502",
      "label": "MMMU",
      "type": "a"
    },
    {
      "href": "https://github.com/lupantech/MathVista",
      "label": "MathVista",
      "type": "a"
    },
    {
      "href": "https://charxiv.github.io/",
      "label": "CharXiv",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2304.02643",
      "label": "Segment Anything Model",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2408.00714",
      "label": "SAM 2",
      "type": "a"
    },
    {
      "href": "https://latent.space/p/sam2",
      "label": "our pod",
      "type": "a"
    },
    {
      "href": "https://github.com/IDEA-Research/GroundingDINO",
      "label": "GroundingDINO",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2304.08485",
      "label": "LLaVA",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/neurips-2023-papers",
      "label": "our pod",
      "type": "a"
    },
    {
      "href": "https://huyenchip.com/2023/10/10/multimodal.html",
      "label": "Flamingo",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2405.09818",
      "label": "Chameleon",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2411.14402",
      "label": "AIMv2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2404.12387",
      "label": "Core",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2022-06-09-vlm/",
      "label": "are at least 4 streams of visual LM work",
      "type": "a"
    },
    {
      "href": "https://cdn.openai.com/papers/GPTV_System_Card.pdf",
      "label": "GPT4V system card",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2309.17421",
      "label": "derivative papers",
      "type": "a"
    },
    {
      "href": "https://blog.roboflow.com/gpt-4o-object-detection/",
      "label": "finetuning 4o vision",
      "type": "a"
    },
    {
      "href": "https://mistral.ai/news/pixtral-large/",
      "label": "Pixtral",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-llama-32-on-device-1b3b-and-multimodal/",
      "label": "Llama 3.2",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=T7sxvrJLJ14",
      "label": "Moondream",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=42505038",
      "label": "QVQ",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=43282905",
      "label": "Mistral OCR",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=43187209",
      "label": "VLM Run",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2212.04356",
      "label": "Whisper",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=33884716",
      "label": "v2",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=38166965",
      "label": "v3",
      "type": "a"
    },
    {
      "href": "https://github.com/huggingface/distil-whisper",
      "label": "distil-whisper",
      "type": "a"
    },
    {
      "href": "https://amgadhasan.substack.com/p/demystifying-openais-new-whisper",
      "label": "v3 Turbo",
      "type": "a"
    },
    {
      "href": "http://AudioPaLM/",
      "label": "AudioPaLM",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2205.04421?utm_source=chatgpt.com",
      "label": "NaturalSpeech",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2403.03100",
      "label": "v3",
      "type": "a"
    },
    {
      "href": "http://Moshi/",
      "label": "Kyutai Moshi",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=hm2IJSKcYvo",
      "label": "high profile demo",
      "type": "a"
    },
    {
      "href": "https://www.hume.ai/blog/introducing-octave",
      "label": "Hume OCTAVE",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/realtime-api",
      "label": "OpenAI Realtime API: The Missing Manual",
      "type": "a"
    },
    {
      "href": "https://www.cartesia.ai/blog/state-of-voice-ai-2024",
      "label": "the State of Voice 2024",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/notebooklm",
      "label": "we got the deepest description of the modeling process",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2006.11477",
      "label": "wav2vec",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2112.10752",
      "label": "Latent Diffusion",
      "type": "a"
    },
    {
      "href": "https://stability.ai/news/stable-diffusion-v2-release",
      "label": "SD2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2307.01952",
      "label": "SDXL",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2403.03206",
      "label": "SD3",
      "type": "a"
    },
    {
      "href": "https://github.com/black-forest-labs/flux",
      "label": "BFL Flux",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2102.12092",
      "label": "DALL-E",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2204.06125",
      "label": "DALL-E-2",
      "type": "a"
    },
    {
      "href": "https://cdn.openai.com/papers/dall-e-3.pdf",
      "label": "DALL-E-3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2205.11487",
      "label": "Imagen",
      "type": "a"
    },
    {
      "href": "https://deepmind.google/technologies/imagen-2/",
      "label": "Imagen 2",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2408.07009",
      "label": "Imagen 3",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2303.01469",
      "label": "Consistency Models",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2310.04378",
      "label": "LCMs",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/tldraw",
      "label": "quick draw viral moment of Dec 2023",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2410.11081",
      "label": "sCMs",
      "type": "a"
    },
    {
      "href": "https://github.com/Tongyi-MAI/Z-Image",
      "label": "DMDs",
      "type": "a"
    },
    {
      "href": "https://openai.com/index/sora/",
      "label": "Sora",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2212.09748",
      "label": "the DiT paper",
      "type": "a"
    },
    {
      "href": "https://artificialanalysis.ai/text-to-video/arena?tab=Leaderboard",
      "label": "open weights competitors",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2412.00131",
      "label": "OpenSora",
      "type": "a"
    },
    {
      "href": "https://lilianweng.github.io/posts/2024-04-12-diffusion-video/",
      "label": "Lilian Weng survey here",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=yJ396Ksiv2s",
      "label": "Autoregressive image generation",
      "type": "a"
    },
    {
      "href": "https://developers.googleblog.com/en/experiment-with-gemini-20-flash-native-image-generation/",
      "label": "Gemini",
      "type": "a"
    },
    {
      "href": "https://news.ycombinator.com/item?id=43474112",
      "label": "4o",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2406.06525",
      "label": "Llama’s",
      "type": "a"
    },
    {
      "href": "https://www.reddit.com/r/singularity/comments/1exsq4d/introducing_ideogram_20_our_most_advanced/",
      "label": "Ideogram",
      "type": "a"
    },
    {
      "href": "https://www.recraft.ai/",
      "label": "Recraft",
      "type": "a"
    },
    {
      "href": "https://buttondown.com/ainews/archive/ainews-halfmoon-is-reve-image-a-new-sota-image/",
      "label": "Reve Image",
      "type": "a"
    },
    {
      "href": "https://github.com/genmoai/mochi",
      "label": "Genmo",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/comfyui",
      "label": "we were first to interview",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=1mG678f1ZYU&pp=ygUOdGV4dCBkaWZmdXNpb24%3D",
      "label": "Text Diffusion",
      "type": "a"
    },
    {
      "href": "https://www.inceptionlabs.ai/blog/mercury-refreshed",
      "label": "Mercury/Inception",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2302.03917",
      "label": "Music Diffusion",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2106.09685",
      "label": "LoRA",
      "type": "a"
    },
    {
      "href": "http://arxiv.org/abs/2305.14314",
      "label": "QLoRA",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/cosine",
      "label": "confirmed on pod",
      "type": "a"
    },
    {
      "href": "https://www.answer.ai/posts/2024-03-06-fsdp-qlora.html",
      "label": "FSDP+QLoRA",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2305.18290",
      "label": "DPO",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/1707.06347",
      "label": "PPO",
      "type": "a"
    },
    {
      "href": "https://platform.openai.com/docs/guides/fine-tuning",
      "label": "Preference Finetuning",
      "type": "a"
    },
    {
      "href": "https://arxiv.org/abs/2404.03592",
      "label": "ReFT",
      "type": "a"
    },
    {
      "href": "https://www.microsoft.com/en-us/research/blog/orca-agentinstruct-agentic-flows-can-be-effective-synthetic-data-generators/",
      "label": "Orca 3/AgentInstruct",
      "type": "a"
    },
    {
      "href": "https://www.latent.space/p/2024-syndata-smolmodels",
      "label": "Synthetic Data picks at NeurIPS",
      "type": "a"
    },
    {
      "href": "https://www.interconnects.ai/p/openais-reinforcement-finetuning",
      "label": "RL Finetuning for o1",
      "type": "a"
    },
    {
      "href": "https://x.com/swyx/status/1867990396762243324",
      "label": "Noam Brown’s many public talks",
      "type": "a"
    },
    {
      "href": "https://github.com/unslothai/unsloth",
      "label": "the Unsloth notebooks",
      "type": "a"
    },
    {
      "href": "https://www.philschmid.de/fine-tune-llms-in-2025",
      "label": "How to fine-tune open LLMs",
      "type": "a"
    }
  ],
  "repair_version": 2,
  "repaired_at": "2026-10-14T10:46:57Z",
  "source_url": "https://www.latent.space/p/2025-papers",
  "papers": [
    {
      "status": "ok",
      "url": "https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf",
      "bytes": 541036,
      "reused": false,
      "local_path": "papers/repair/a3e86ad05f37e40d21b6.pdf"
    },
    {
      "status": "ok",
      "url": "https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf",
      "bytes": 582775,
      "reused": false,
      "local_path": "papers/repair/3c62359064baa28a7743.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2005.14165",
      "bytes": 6768044,
      "reused": true,
      "local_path": "papers/repair/0c1629d8d07d676c2aae.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2107.03374",
      "bytes": 1885177,
      "reused": false,
      "local_path": "papers/repair/c41850b212e88f282029.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2203.02155",
      "bytes": 1797405,
      "reused": true,
      "local_path": "papers/repair/4b8cf107df7118e0ffe8.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2303.08774",
      "bytes": 5245564,
      "reused": false,
      "local_path": "papers/repair/e00f74df20cd8064349a.pdf"
    },
    {
      "status": "ok",
      "url": "https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf",
      "bytes": 28241567,
      "reused": false,
      "local_path": "papers/repair/1b2bfa31a560671d4b8e.pdf"
    },
    {
      "status": "ok",
      "url": "https://www-cdn.anthropic.com/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf?s=09",
      "bytes": 4876911,
      "reused": false,
      "local_path": "papers/repair/aa134c8b57f85ce1dc59.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2312.11805",
      "bytes": 27484659,
      "reused": false,
      "local_path": "papers/repair/b515eb2177e7e678c142.pdf"
    },
    {
      "status": "ok",
      "url": "https://storage.googleapis.com/deepmind-media/gemini/gemini_v2_5_report.pdf",
      "bytes": 10130715,
      "reused": false,
      "local_path": "papers/repair/fafd703f15bda2e9dce8.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2408.00118",
      "bytes": 610320,
      "reused": true,
      "local_path": "papers/repair/4de2cdfb0c42fc1d6aa3.pdf"
    },
    {
      "status": "ok",
      "url": "https://storage.googleapis.com/deepmind-media/gemma/Gemma3Report.pdf",
      "bytes": 3699196,
      "reused": false,
      "local_path": "papers/repair/1a935e7ee2116628b1e0.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2302.13971",
      "bytes": 726566,
      "reused": false,
      "local_path": "papers/repair/c0170963c7095dc7febf.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2307.09288",
      "bytes": 13661300,
      "reused": false,
      "local_path": "papers/repair/01c99984e3c41b2873f5.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2407.21783",
      "bytes": 9833173,
      "reused": true,
      "local_path": "papers/repair/213e617e0eb2b87ce1fa.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.06825",
      "bytes": 3749788,
      "reused": true,
      "local_path": "papers/repair/0aa8d5bd14ffe7b6d4d5.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.04088",
      "bytes": 2475990,
      "reused": true,
      "local_path": "papers/repair/041bf1008157bd282f03.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2410.07073",
      "bytes": 12640953,
      "reused": false,
      "local_path": "papers/repair/1fd34455e8d347fa5a85.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2505.09388",
      "bytes": 779424,
      "reused": false,
      "local_path": "papers/repair/c08fa7046aa3f4744a1a.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.02954",
      "bytes": 7474751,
      "reused": false,
      "local_path": "papers/repair/b3ba4aaa28f20063720a.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.14196",
      "bytes": 1911399,
      "reused": false,
      "local_path": "papers/repair/e14e595a1deca88549df.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2402.03300",
      "bytes": 1844642,
      "reused": true,
      "local_path": "papers/repair/dd0de0096b51ee9b7ed5.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.06066",
      "bytes": 733666,
      "reused": false,
      "local_path": "papers/repair/76028dd64c1825572894.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2405.04434",
      "bytes": 1586880,
      "reused": false,
      "local_path": "papers/repair/5175a77604cd16ff9e0e.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2502.19587",
      "bytes": 857621,
      "reused": false,
      "local_path": "papers/repair/6b068cc9f539adb6b578.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2402.00838",
      "bytes": 562224,
      "reused": true,
      "local_path": "papers/repair/8fd8880ac3f82766be24.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2409.17146",
      "bytes": 6576158,
      "reused": true,
      "local_path": "papers/repair/b56ffcb14cd1220cae1b.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2409.02060",
      "bytes": 6501620,
      "reused": false,
      "local_path": "papers/repair/c3dab3d3cf6c6b225fad.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2412.08905",
      "bytes": 1016588,
      "reused": false,
      "local_path": "papers/repair/a0b4fa22aff851d502fe.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2407.21075",
      "bytes": 20588337,
      "reused": true,
      "local_path": "papers/repair/e3858449ae351a2bc258.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2001.08361",
      "bytes": 2493004,
      "reused": false,
      "local_path": "papers/repair/9d7d7113c60c89afd004.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2203.15556",
      "bytes": 6004349,
      "reused": true,
      "local_path": "papers/repair/70766cef482201671d90.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2206.07682",
      "bytes": 842905,
      "reused": false,
      "local_path": "papers/repair/a904af74a10465c0b699.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2304.15004",
      "bytes": 1894670,
      "reused": false,
      "local_path": "papers/repair/162161e6c9b97cb4cc03.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.00448",
      "bytes": 3957782,
      "reused": false,
      "local_path": "papers/repair/316f8fc3cc393710ba29.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2502.16982",
      "bytes": 2077505,
      "reused": false,
      "local_path": "papers/repair/2abc387082032ecc3e49.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2305.20050",
      "bytes": 4384742,
      "reused": false,
      "local_path": "papers/repair/16e5e83b6544701468c6.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2501.19393",
      "bytes": 1480533,
      "reused": false,
      "local_path": "papers/repair/e43041c5d25b74028f35.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2009.03300",
      "bytes": 2074939,
      "reused": true,
      "local_path": "papers/repair/eabe88a3d618ccd5d1e4.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2311.12022",
      "bytes": 617128,
      "reused": true,
      "local_path": "papers/repair/9aae134cc3cb0f937388.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2206.04615",
      "bytes": 3874852,
      "reused": false,
      "local_path": "papers/repair/2e24c8377bfad96bd249.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2406.01574",
      "bytes": 959520,
      "reused": true,
      "local_path": "papers/repair/0631cba10262c7677dc1.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2210.09261",
      "bytes": 5375941,
      "reused": false,
      "local_path": "papers/repair/46ed07e5c491b3fb4081.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.16049",
      "bytes": 1088520,
      "reused": false,
      "local_path": "papers/repair/70bc8fcf6eb554c818ae.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2412.15204",
      "bytes": 9951437,
      "reused": false,
      "local_path": "papers/repair/c415bedae6d5bdbea511.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2406.10149",
      "bytes": 5321327,
      "reused": false,
      "local_path": "papers/repair/92ccea2b3b77a6d9c105.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2103.03874",
      "bytes": 2030715,
      "reused": false,
      "local_path": "papers/repair/61e56e93685b60f6cb83.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2411.04872",
      "bytes": 1687781,
      "reused": false,
      "local_path": "papers/repair/1a64c0deb99e83fa07e2.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2311.07911",
      "bytes": 464152,
      "reused": false,
      "local_path": "papers/repair/aa7d0c8c51f0f0ea8fdd.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2410.15553",
      "bytes": 1250137,
      "reused": false,
      "local_path": "papers/repair/11840c4db4cd7a408247.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2501.17399",
      "bytes": 2789352,
      "reused": false,
      "local_path": "papers/repair/f24f5ba7bdb8d2fbb7fd.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2406.06608",
      "bytes": 3176422,
      "reused": false,
      "local_path": "papers/repair/f2cdbc59d4e9f6c4190f.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2201.11903",
      "bytes": 891773,
      "reused": true,
      "local_path": "papers/repair/52cc4c4f6cf368123b97.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2112.00114",
      "bytes": 408605,
      "reused": false,
      "local_path": "papers/repair/790393059a76b0c7420e.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2205.11916",
      "bytes": 762445,
      "reused": false,
      "local_path": "papers/repair/7ee8d7eb1358c6773ecd.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2305.10601",
      "bytes": 799685,
      "reused": false,
      "local_path": "papers/repair/7b69e34e7c5bdef7b2b4.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2101.00190",
      "bytes": 1569113,
      "reused": false,
      "local_path": "papers/repair/a704430cffe77134ac6b.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2402.10200",
      "bytes": 1288337,
      "reused": false,
      "local_path": "papers/repair/be1144c6c92c161e675f.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2211.01910",
      "bytes": 4085477,
      "reused": false,
      "local_path": "papers/repair/8f60448c0d6b2d6e9d1d.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.03714",
      "bytes": 460814,
      "reused": false,
      "local_path": "papers/repair/b9a313a53a91d8964133.pdf"
    },
    {
      "status": "ok",
      "url": "https://dennyzhou.github.io/LLM-Reasoning-Stanford-CS-25.pdf",
      "bytes": 1410666,
      "reused": false,
      "local_path": "papers/repair/86adbdeab6ba4ed60b37.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/1603.09320",
      "bytes": 2636958,
      "reused": false,
      "local_path": "papers/repair/7f2db157afe85482dc33.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2005.11401",
      "bytes": 885323,
      "reused": false,
      "local_path": "papers/repair/fc8d32f196e0f0c27ffb.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2210.07316",
      "bytes": 769075,
      "reused": false,
      "local_path": "papers/repair/c1b726e030debeebfc42.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2404.16130",
      "bytes": 6893854,
      "reused": false,
      "local_path": "papers/repair/b101ca0e7d5283641404.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2309.15217",
      "bytes": 231984,
      "reused": false,
      "local_path": "papers/repair/060dc88922a3d3638629.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2407.07858v1",
      "bytes": 1165789,
      "reused": false,
      "local_path": "papers/repair/0ea7f54efc8af9ab8d1f.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2407.16833",
      "bytes": 657247,
      "reused": false,
      "local_path": "papers/repair/ebe21d314c0082cfb551.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.06770",
      "bytes": 4533722,
      "reused": true,
      "local_path": "papers/repair/500a09d951fecb01ca85.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2405.15793",
      "bytes": 4963538,
      "reused": false,
      "local_path": "papers/repair/4496a0fcb05844f38032.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2410.03859",
      "bytes": 4379827,
      "reused": true,
      "local_path": "papers/repair/2cb52e4c097eb08ba577.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2406.12045",
      "bytes": 663224,
      "reused": true,
      "local_path": "papers/repair/b2b3f07dc2060b39abb0.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2311.12983",
      "bytes": 4254059,
      "reused": true,
      "local_path": "papers/repair/7d67f676a7c6219c7bef.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2210.03629",
      "bytes": 633805,
      "reused": true,
      "local_path": "papers/repair/c08ee62f780204fcfff6.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2302.04761",
      "bytes": 657966,
      "reused": false,
      "local_path": "papers/repair/68512ec86fa6106aa473.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2303.17580",
      "bytes": 3417598,
      "reused": false,
      "local_path": "papers/repair/d88495d43010cc38191c.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.08560",
      "bytes": 663708,
      "reused": false,
      "local_path": "papers/repair/79f95208d786cc7e6610.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2308.00352",
      "bytes": 16753634,
      "reused": false,
      "local_path": "papers/repair/711305ee1710b630c5e0.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2308.08155",
      "bytes": 3571809,
      "reused": false,
      "local_path": "papers/repair/f7b4b3114d3d299e9e7e.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2305.16291",
      "bytes": 18830859,
      "reused": false,
      "local_path": "papers/repair/1b7fd1c26a720921b57c.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2309.02427",
      "bytes": 2785564,
      "reused": false,
      "local_path": "papers/repair/0f608fb22099fe2548f1.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2409.07429",
      "bytes": 941281,
      "reused": false,
      "local_path": "papers/repair/537c9d6228739b703d56.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2211.15533",
      "bytes": 736600,
      "reused": false,
      "local_path": "papers/repair/776366151a7585761fe7.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2402.19173",
      "bytes": 1114300,
      "reused": false,
      "local_path": "papers/repair/35d74f78274ef91ea152.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2409.12186",
      "bytes": 1204247,
      "reused": false,
      "local_path": "papers/repair/b50107334b0f8260a5b0.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2312.02143",
      "bytes": 1907961,
      "reused": false,
      "local_path": "papers/repair/9bb58c5c90f06c74fd05.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2502.06807v1",
      "bytes": 1236874,
      "reused": false,
      "local_path": "papers/repair/9195be044835d7ece019.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.08500",
      "bytes": 475308,
      "reused": true,
      "local_path": "papers/repair/67008eeb943c437e182e.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2412.15004v1",
      "bytes": 1019258,
      "reused": false,
      "local_path": "papers/repair/0ba8df3762966c20909e.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/1506.02640",
      "bytes": 5296750,
      "reused": false,
      "local_path": "papers/repair/27e378fce20ee7c0b279.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2304.08069",
      "bytes": 2995096,
      "reused": false,
      "local_path": "papers/repair/fbd3513cd1dc2699af9d.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2103.00020",
      "bytes": 6813639,
      "reused": false,
      "local_path": "papers/repair/7db0b9244bb1fd42ae98.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2010.11929",
      "bytes": 3743814,
      "reused": true,
      "local_path": "papers/repair/81153ab91885fd77745c.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2201.12086",
      "bytes": 1303121,
      "reused": false,
      "local_path": "papers/repair/9f86c994dce24bfd0fe7.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2301.12597",
      "bytes": 6737540,
      "reused": false,
      "local_path": "papers/repair/550756db727a4887cce7.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2401.06209",
      "bytes": 2615753,
      "reused": false,
      "local_path": "papers/repair/310816ba9be7bc170fd9.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2311.16502",
      "bytes": 16874947,
      "reused": false,
      "local_path": "papers/repair/57258dc74a43fd190d31.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2304.02643",
      "bytes": 15644599,
      "reused": false,
      "local_path": "papers/repair/f6782c6c60b1d53c35dd.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2408.00714",
      "bytes": 12709657,
      "reused": false,
      "local_path": "papers/repair/794bdf5c141ff278d61c.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2304.08485",
      "bytes": 5906135,
      "reused": false,
      "local_path": "papers/repair/11628fe93e09857fc735.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2405.09818",
      "bytes": 25245551,
      "reused": false,
      "local_path": "papers/repair/28592f7cd7b51dc91733.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2411.14402",
      "bytes": 4788339,
      "reused": false,
      "local_path": "papers/repair/18d6576734a9ac2e1b33.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2404.12387",
      "bytes": 3817001,
      "reused": false,
      "local_path": "papers/repair/254f058fcb6e5e531005.pdf"
    },
    {
      "status": "ok",
      "url": "https://cdn.openai.com/papers/GPTV_System_Card.pdf",
      "bytes": 2956129,
      "reused": false,
      "local_path": "papers/repair/0ff4973faf1157037cf2.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2309.17421",
      "bytes": 45669728,
      "reused": false,
      "local_path": "papers/repair/7ee97e26202c36e350b4.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2212.04356",
      "bytes": 1008644,
      "reused": false,
      "local_path": "papers/repair/f46bfeac76b09a554dac.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2205.04421",
      "bytes": 710525,
      "reused": false,
      "local_path": "papers/repair/7a3bbd8411dedc0cbb9b.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2403.03100",
      "bytes": 745607,
      "reused": false,
      "local_path": "papers/repair/fc166396401cc671f376.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2006.11477",
      "bytes": 740863,
      "reused": false,
      "local_path": "papers/repair/13f1c73bdef7a5028210.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2112.10752",
      "bytes": 40842586,
      "reused": false,
      "local_path": "papers/repair/84de7ccc1cc12ec4cb29.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2307.01952",
      "bytes": 18407785,
      "reused": false,
      "local_path": "papers/repair/44aca15089da3be9681d.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2403.03206",
      "bytes": 33973124,
      "reused": false,
      "local_path": "papers/repair/ebe7cc58d51134dfd1f0.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2102.12092",
      "bytes": 10126551,
      "reused": false,
      "local_path": "papers/repair/2a89049babb7ba4cc73c.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2204.06125",
      "bytes": 42905303,
      "reused": false,
      "local_path": "papers/repair/037740aca2c3ab83f801.pdf"
    },
    {
      "status": "ok",
      "url": "https://cdn.openai.com/papers/dall-e-3.pdf",
      "bytes": 28086764,
      "reused": false,
      "local_path": "papers/repair/59f33a7592ea074b14a5.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2205.11487",
      "bytes": 11365725,
      "reused": false,
      "local_path": "papers/repair/8c48830dcf885dd9be79.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2408.07009",
      "bytes": 31513597,
      "reused": false,
      "local_path": "papers/repair/b7bc2f204907ffdf4614.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2303.01469",
      "bytes": 31529980,
      "reused": false,
      "local_path": "papers/repair/c129e904a22d86f9e0a4.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2310.04378",
      "bytes": 15939992,
      "reused": false,
      "local_path": "papers/repair/3366ac519f420b807343.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2410.11081",
      "bytes": 29354775,
      "reused": false,
      "local_path": "papers/repair/ea319faa4f00db0e5037.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2212.09748",
      "bytes": 43824062,
      "reused": false,
      "local_path": "papers/repair/6ff6a352b0138ad5d782.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2412.00131",
      "bytes": 11561874,
      "reused": false,
      "local_path": "papers/repair/fd9d8c642a6725b903eb.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2406.06525",
      "bytes": 21965938,
      "reused": true,
      "local_path": "papers/repair/7947c34b3f7246e1a0ef.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2302.03917",
      "bytes": 609842,
      "reused": false,
      "local_path": "papers/repair/9499cdad2f23479bb813.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2106.09685",
      "bytes": 1609513,
      "reused": true,
      "local_path": "papers/repair/e7d2fed957472d7677ec.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2305.14314",
      "bytes": 1065470,
      "reused": true,
      "local_path": "papers/repair/df2f754581ddf63a5d62.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2305.18290",
      "bytes": 1299212,
      "reused": true,
      "local_path": "papers/repair/ba63f83d413059cb8722.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/1707.06347",
      "bytes": 2923532,
      "reused": true,
      "local_path": "papers/repair/5c73fa759d96331836bf.pdf"
    },
    {
      "status": "ok",
      "url": "https://arxiv.org/pdf/2404.03592",
      "bytes": 2329282,
      "reused": false,
      "local_path": "papers/repair/c36edd583694ec9debc1.pdf"
    }
  ],
  "actual_video_files": [
    {
      "local_path": "attachments/fd41a5f828b9/videos/hnIMY9pLPdg.mp4",
      "bytes": 460676566,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "3726.021000",
      "status": "ok"
    }
  ],
  "body_chars": 14788,
  "expected_body_images": 1,
  "discovery_status": "ok",
  "delivery_status": "partial",
  "missing_reason": "no_1080_or_720",
  "raw_html": "attachments/fd41a5f828b9/pages/source.html",
  "local_html": "attachments/fd41a5f828b9/pages/article.html",
  "body_base_path": "attachments/fd41a5f828b9/pages/",
  "body_markdown_path": "attachments/fd41a5f828b9/pages/article.md"
}