WOVEN Teaches MLLMs Visual Transition Reasoning That Transfers Across Tasks

WOVEN trains MLLMs on world modeling as state transitions, focusing on visual transition reasoning over (s,a,s′). The shared primitive transfers across tasks, improving results on 22 of 26 benchmarks.

2026-10-10 ~ 2026-10-10 · 2 related posts

• WOVEN: visual transition reasoning as a shared primitive lifts 22 of 26 benchmarks by up to 27.3 points — mohitban47 · 2026-10-10
• WOVEN shows visual transition reasoning is a teachable primitive that transfers across tasks — mohitban47 · 2026-10-10