Faithful-MR1 improves faithful visual perception and reasoning in MLLMs by anchoring a <Focus> token on image regions before reasoning and reinforcing correct attention via counterfactual rewards, outperforming baselines on Qwen2.5-VL models with less data.
The height of the cone is the opposite side to the 60-degree angle, and the slant height is the hypotenuse
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Faithful-MR1: Faithful Multimodal Reasoning via Anchoring and Reinforcing Visual Attention
Faithful-MR1 improves faithful visual perception and reasoning in MLLMs by anchoring a <Focus> token on image regions before reasoning and reinforcing correct attention via counterfactual rewards, outperforming baselines on Qwen2.5-VL models with less data.