Evidence-RL：邁向證據密集型的視覺推理

摘要

視覺語言模型（VLM）應基於具體的圖像證據來回答，而非語言先驗、數據集捷徑或無關的視覺上下文。現有的感知感知後訓練方法通過全局擾動或注意力代理來鼓勵模型使用圖像，但它們並未測試所取樣答案是否因果依賴於支持它的局部證據。我們提出反事實證據解耦（CED），這是一種針對 VLM 接地（grounding）的訓練時證據審計機制。對於每個回應，CED 會中和一個以物體為中心的證據區域，並將由此產生的支持度下降與匹配的非證據區域進行比較。我們將此信號與答案正確性結合到 GRPO 中，獎勵依賴證據路徑而非捷徑或干擾路徑的正確答案。CED 使用弱物體級提案，無需針對問題的證據標註，且不增加推理時間開銷。在九個公開基準測試和四個骨幹網路中，CED 優於先前的基於強化學習的後訓練方法，且針對性分析驗證了其以物體為中心的信號。

English

Vision-Language Models (VLMs) should answer from concrete image evidence rather than language priors, dataset shortcuts, or irrelevant visual context. Existing perception-aware post-training methods encourage image use through global perturbations or attention proxies, but they do not test whether a sampled answer causally depends on the local evidence that supports it. We propose Counterfactual Evidence Disentanglement (CED), a training-time evidence audit for VLM grounding. For each response, CED neutralizes an object-centric Evidence Region and compares the resulting support drop against matched non-evidence Regions. We combine this signal with answer correctness inside GRPO, rewarding correct answers that rely on the evidence path rather than shortcut or nuisance paths. CED uses weak object-level proposals, requires no question-specific evidence annotations, and adds no inference-time overhead. Across nine public benchmarks and four backbones, CED outperforms prior RL-based post-training methods, with targeted analyses verifying its object-centric signal.