🔥Telegram İfşa

Check us out also on bsky.app/profile/krishn…!

ICLR is a top ML conference. All 10k+ papers from 2025 are in Open Review. The top rated papers include: — Scaling LLM interpretability to GPT4 scale — Changing light source w consistent image — 100x faster diffusion models — A provable theory for LLM jailbreaks Thread... 1/10

Causal considerations can determine the utility of machine learning assisted GWAS biorxiv.org/cgi/content/sh… #biorxiv_genetic

HypotheSAEs outperforms strong LLM baselines, generates new discoveries even on well-studied datasets, and comes with easy-to-use code. We hope this will be helpful not just to CS folks, but to many in social/health sciences - please consider resharing to help reach them.

Using simulations & RWD, Sumit Mukherjee shows ML-imputed phenotypes boost GWAS power only when built from upstream biomarkers. Downstream proxies inflate FDR, & high predictive R2 can mislead—genetic vs. environ. correlation matters. Pick proxies w/ causal insight! #CHIL2025papers

I’m honored to join Fred Hutch as Professor and Program Head of Biostatistics, and as the Donald and Janet K. Guthrie Endowed Chair in Statistics. Excited to be part of a deeply collaborative and scientifically vibrant community with a rich legacy of impact.

Microsoft claims their new AI framework diagnoses 4x better than doctors. I'm a medical doctor and I actually read the paper. Here's my perspective on why this is both impressive AND misleading ... 🧵

Q&A: Transparency in medical AI systems is vital, UW researchers say washington.edu/news/2025/09/1…