RAPO:長鏈多模態推理中的傳播感知視覺保留反思錨點
Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning
RAPO 透過識別高熵推理位置作為稀疏反思錨點,並以 chain-masked KL 目標訓練策略在後續推理中保留視覺影響。在 Qwen3-VL-2B-Instruct 上,RAPO 將六項基準的平均 mean@8 accuracy 從 GRPO 的 40.79 提升至 43.01。標準推理時無須額外重新注入視覺資訊或選擇錨點。
RAPO: Keeping Visual Evidence in the Loop during Long-Chain Reasoning
Excited to share our NeurIPS 2026 paper, Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning!
As multimodal reasoning chains grow longer, models can gradually lose their dependence on visual evidence and rely increasingly on generated text.
We introduce Reflection-Anchor Policy Optimization (RAPO), which identifies high-entropy reasoning positions as sparse reflection anchors and trains the policy to preserve visual influence in downstream reasoning through a chain-masked KL objective.
Key highlights:
- Principled motivation: An information-theoretic analysis of local branching and downstream visual propagation.
- Propagation-aware optimization: Strengthening visual dependence beyond individual reasoning steps.
- Stronger performance: On Qwen3-VL-2B-Instruct, RAPO improves the six-benchmark average from 40.79 (GRPO) to 43.01 (mean@8 accuracy).
- Standard inference: No additional visual re-injection or anchor selection during decoding.
Paper: https://arxiv.org/abs/2605.09614
Project: https://coder-gx.github.io/RAPO/
GitHub: https://github.com/coder-gx/RAPO
We welcome feedback and discussion!
來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co