保留正文与补采附件。源页面当前未重新核验；具体限制见交付清单。
SubscribeSign in Home Tabular Foundation Models Archive About
LatestTopDiscussions
Trends in tabular foundation research
Based on 151 papers from ICML workshop "Foundation Models for Structured Data"
Jul 22 • Christoph Molnar
28
3
Time for a change
After 4 years of writing, I'm exploring what's next for me in ML / data science / AI (or whatever you want to call it)
Jul 16 • Christoph Molnar
32
8
TabFM minus the hype
top performance on TabArena; large model; slow inference; non-commercial license
Jul 7 • Christoph Molnar
34
Which ML models produce the best quantile estimates?
Results from ScoringBench
Jun 30 • Christoph Molnar
14
2
When trees still beat tabular foundation models
As you might have noticed, I’m rather optimistic about tabular foundation models.
Jun 23 • Christoph Molnar
30
8
1
What is TabPFN's Thinking mode?
When I read the TabPFN-3 Technical Report, the benchmark for “TabPFN-3-Thinking” stood out: it appeared at the top of the TabArena benchmark for large…
Jun 16 • Christoph Molnar
16
1
2
How TabICL and TabPFN handle missing values
You can give training or test data with missing values to the tabular foundation models TabPFN and TabICL, and the prediction will “just work”.
Jun 9 • Christoph Molnar
23
1
3
TabPFN and TabICL are not everything
An overview of other families for tabular foundation models
Jun 2 • Christoph Molnar
22
2
I’m writing a book on Tabular Foundation Models
tl;dr: In-progress book here: tabularfoundationmodels.com
May 26 • Christoph Molnar
33
12
2
Tabular ML is entering a new benchmark era
From static and narrow benchmarks to live, capability-driven evaluation
May 19 • Christoph Molnar
16
How to make Tabular Foundation Model inference faster
GPU, KV caching, context optimization, and other inference tricks
May 12 • Christoph Molnar
21
2
1
Time series forecasting with tabular foundation models
This post is a quick primer on using tabular foundation models for time series forecasting.
May 5 • Christoph Molnar
19
4
3
Train or Test for Feature Effect Estimation? We Finally Have an Answer
A guest post by XAI researcher Timo Heiß
Apr 28 • Christoph Molnar
16
3
3
Context is the new training
Tabular foundation models replace retraining with editable inference-time data
Apr 21 • Christoph Molnar
39
1
3
Regression should predict full distributions
A "hidden" feature of tabular foundation models
Apr 14 • Christoph Molnar
39
2
1
TabArena explained
What chess ratings tell us about ML models
Mar 31 • Christoph Molnar
20
4
1
The interpretability tax on tabular foundation models
Model-agnostic interpretability for tabular foundation models
Mar 24 • Christoph Molnar
25
9
1
I’m betting on tabular foundation models
The shipping container moment of data science
Mar 17 • Christoph Molnar
66
9
9
The Random Forest of the 2030s?
Three scenarios for the future of tabular ML, plus one that makes them all irrelevant
Mar 3 • Christoph Molnar
24
2
A regression example with TabICLv2
Strong performance at a steep interpretability cost (at least on CPU)
Feb 24 • Christoph Molnar
27
3
1
The state of Tabular Foundation Models (2026)
Pre-trained models, startups, and recommendation.
Feb 17 • Christoph Molnar
41
7
4
How Tabular Foundation Models learn without real data
The pre-training recipe behind TFMs
Feb 10 • Christoph Molnar
36
2
4
The architecture behind TabPFN
How transformers learn from entire tables
Feb 3 • Christoph Molnar
43
7
1
How PFNs make tabular foundation models work
Bayesian logic, synthetic priors, and in-context learning
Jan 21 • Christoph Molnar
62
2
4
© 2026 Christoph Molnar · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture