Kundaje: scFMs trained on observational data cannot learn causal effects reliably
anshulkundaje · x · 2026-10-06
In post 4 of his critique thread, Anshul Kundaje states his central methodological point: scFMs trained on observational data do not and cannot learn causal effects in any reliable fashion, and thus can't be used directly for perturbation prediction or causal discovery. Context from post 3: the paper only shows fine-tuned scFMs beating trivial baselines under updated metrics, which doesn't vindicate the scFM route.
More from Research
- User Sim Index is broken: trivial bot scores 95% across behavioral dims — ericzelikman · 2026-10-06
- Used OpenAI Dots as a Free Agent Swarm to Break a 47-Year-Old Math Record — jaxchang · 2026-10-06
- Stanford prof says current experiment designs can't train causal virtual cell models — anshulkundaje · 2026-10-06
- TIDES dataset on multi-party and multi-agent collaboration to debut at COLM 2026 — josephseering · 2026-10-06
- LakeQuest QA benchmark, testing RAG on messy enterprise tables and docs, hits COLM — hllo_wrld · 2026-10-06
- Frontier Data Summit lineup: Chollet, Dawn Song headline batch of new agent benchmarks — StanfordAILab · 2026-10-06