Stanford prof questions whether Arc Institute cell-sim contest winner truly generalizes
anshulkundaje · x · 2026-11-20
Stanford professor Anshul Kundaje seriously questions whether the Arc Institute cell-simulation contest's winning solution is truly generalizable, or just a nearest-neighbor retrieval that works because test scenarios happen to be close to training ones. Defenders counter that the test cell lines included one close, one partial, and one never-seen type, and the data used completely different Illumina single-cell technology, validating cross-lab reproducibility.
More from Research
- MIT's Hiroshi Ishii shares Tangible Media Group's Creative Tension principles — ishii_mit · 2026-11-20
- Train a real LLM entirely in your browser with this interactive demo — morgymcg · 2026-11-20
- Ex-OpenAI Researcher: English Is the New ML Programming Language, Lean the Type System — lukaszkaiser · 2026-11-20
- S1CogBench probes System One decision models across 24 cognitive tasks — Dengzhe Hou · 2026-11-20
- AFFECT from KAIST makes world models dynamics-faithful via action effect matching — John Won · 2026-11-20
- Mind2Act World: a 180-task benchmark exposing robots' mind-action composition failures — Mind2Act-World · 2026-11-20