ReState combines two mechanisms.
Iterative reasoning states help multimodal models refine their own answers
ReState trains one multimodal model to construct, preserve and correct reasoning through repeated updates to a shared state.
Chinese Tech
Chuang Yu · Xuefei Ma · Yuanhao Feng · Wuxiong Zeng · Bo Wang · Minghong Cai · +6 more
MMLab, CUHK · PolyU · Tencent · Peking University · CPII
Research Digest··2 min read
Yu and colleagues recast multimodal chain-of-thought reasoning as an iterative state-update process rather than a sequence of separately trained stages.
Why this paper
From Tencent and 4 others
In one line
ReState reformulates multimodal reasoning as dynamic state transitions, unifying construction, correction, and preservation for improved chain-of-thought.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§