TabPFN-3.5 sets new SOTA on tabular benchmarks, extends to multimodal tables
burny_tech · x · 2026-09-20
A large team including Frank Hutter and Yann LeCun released the TabPFN-3.5 technical report. The tabular foundation model sets a new state of the art on tabular prediction benchmarks and expands to messy real-world data: non-i.i.d. splits, high-cardinality features, wide tables, and tables containing text and images. For engineers building data pipelines and analytical agents, an out-of-the-box model that skips feature engineering and GBDT tuning could radically simplify tabular tool-use and downstream automation.
More from Research
- Researcher urges papers to show actual training data samples, cites his CVPR24 practice — gabriberton · 2026-09-20
- Vernor Vinge's 1993 "Technological Singularity" essay: superhuman AI within 30 years — akbirthko · 2026-09-20
- Self-Rewarding LLMs isn't news: RLAIF has long been standard practice — burny_tech · 2026-09-20
- Long-standing Catalan constant irrationality proof posted, claimed to be LLM-assisted — burny_tech · 2026-09-20
- Brain Runs on 20 Watts: Can Neuromorphic Computing Make AI Less Power-Hungry? — burny_tech · 2026-09-20
- Amid the AI math proof debate, a curated set on proofs across generations — RexDouglass · 2026-09-20