
Currently @MSRNE , ML4Bio PhD Student @ucla | playing with multimodal latent representations and loss functions | prev @genentech
San Francisco
Joined February 2024
- Tweets20
- Following94
- Followers26
- Likes137
Gonna be presenting this (poster) at #ICML26 this Thursday@2:30p!
Itβll be at Hall A #2805
Swing by if youβd like learn more limitations of tabular model evals or just wanna chat :)
Getting scooped on my own paper drop by @DamienTeney π
But he's spot on - tabular language models have been overblown
Reported performance gains often donβt come from emergent generalization at all.
arxiv.org/abs/2602.04031
2
94
Adi Gorla @ ICML π°π· retweeted
The Illusion of Generalization: Re-examining Tabular Language Model Evaluation
Aditya Gorla, @ratishsp
tl;dr: @tunguz and other Kaggle GMs were right all along about tabular DL models.
arxiv.org/abs/2602.04031
Getting scooped on my own paper drop by @DamienTeney π
But he's spot on - tabular language models have been overblown
Reported performance gains often donβt come from emergent generalization at all.
arxiv.org/abs/2602.04031
Some major findings:
1) 39% of datasets show zero improvement over majority class baseline
2) Top performers severely contaminated; complete train-test overlap that evades deduplication
3) Much of the gain comes from instruction tuning alone
So whats actually being learned?
1
118
Most "learning" is either predicting majority class, memorizing contaminated data, or generic instruction-following. Not tabular-specific at all.
Benchmarks need baselines, better dedup, task stratification, and cleaner eval datasets.
Really fun working with @ratishsp on this!
107
Our perspective on heritable confounders in big GWAS meta-analyses on psychiatric disorders is now online @NatureGenet: nature.com/articles/s41588-0β¦, with @andywdahl, @jonathan_flint1, Richard Border et al. Small thread for context 1/n
Some exciting news: Our method CACTI will be a β¨spotlight β¨poster at #ICML2025!
CACTI solves the missing data problem in tabular datasets with a transformer that learns from REAL missingness patterns + semantic context, beating SOTA by 13.4% in the most realistic settings
π§΅1/9
1
2
182
Come check out poster at #icml2025 in East Exhibition Hall A-B: E-1300 at the 11am session on 16/07 : icml.cc/virtual/2025/poster/β¦
Or read our preprint for more details: arxiv.org/abs/2506.02306
8/9
1
171
This work would not have been possible without my amazing collaborators Ryan Wang @ZhengtongLiu @ulzeean and advisor @sr_sankararaman!
9/9
106







