NVIDIA/MIT's Physis-Lang tops Physics-IQ leaderboard, +5.5 points via evolved captions
songhan_mit · x · 2026-09-30
NVIDIA, MIT and Oxford propose Physis-Lang, a self-evolving language representation for video world models. Captions are decomposed into cause, physical law and effect, and a physics-aware critic plus an agent iteratively refine shared captioning guidelines while the captioner stays fixed. The same representation spans data curation, training and inference. On Physics-IQ Verified it takes #1 and #2: Cosmos3-Super scores 48.2, beating the previous leader by 5.5 points; Cosmos3-Nano scores 43.3, surpassing the original Cosmos3-Super.
Related event: NVIDIA Open-Sources Physis-Lang to Teach Video Models Physics(3 posts)→
More from Multimodal
- Argil soft-launches AI storytelling studio, cutting 30s video production from 3 hours to 10 minutes — kevinnbass · 2026-09-30
- Plenio 0.3 ships: local AI music production in ComfyUI with score editor, stems and mastering — Vivid_Promise1700 · 2026-09-30
- NVIDIA's Physis-Lang puts physics reasoning in captions, tops Physics-IQ with Cosmos 3 — NVIDIAAI · 2026-09-30
- One image + one mp3: Opus 5.5 renders a 20-second video scene in 30 minutes — justin_hart · 2026-09-30
- This AI video is so realistic the author's grandma asked if it was AI — techhalla · 2026-09-30
- ComfyUI seeks testers for new Asset System: indexed files, persistent output history — Lexius2129 · 2026-09-30