DiffWAM
World-action model from Zhejiang University, Fudan University, Zhongguancun Academy and CASIA that learns from the difference between consecutive frames. It comes in two variants: DiffWAM-S for small datasets and DiffWAM-L, which trains all modules, for large ones. Code is announced but not released.
Zhejiang University et al. · Nov 15, 2026 · World-action · built on Fast-WAMReport an error
98.8%LIBERO · #17 of 83
2Benchmarks reported
Nov 15, 2026Released
UnknownWeights
Benchmark results2
- 92.7%, self-reported by the authors, not a leaderboard entryPaper / model card ↗Nov 2026· self-reported, 50 clean + 500 randomized demos· n = 10000· non-standard
92.7%, self-reported by the authors, not a leaderboard entrynot ranked
- Paper / model card ↗Nov 2026· n = 4000
#17 of 8398.8%, self-reported by the authors, not a leaderboard entry
Not reported on 13 benchmarks
Wrong or missing number? Report an error