GRIT 训练框架针对 VLM 视觉错位推理问题，并发布 GRIT-Bench
GRIT: Gradient Masking for Visually Misgrounded Tokens in Text-Rich Reasoning
视觉语言模型经常瞎猜图像也能答对题，这篇论文用注意力分析定位了这些 token，再用梯度掩码修掉，还配了个 GRIT-Bench 基准。
论文提出 CVMR 现象：视觉语言模型在中间步骤误读图像信息，最终答案仍可能正确，而标准 RLVR 只看最终答案对错，无法区分这类推理。研究者发现 CVMR token 在注意力模式上与正确对齐的 token 有系统性差异，据此提出 GRIT，用梯度掩码在 RLVR 中引入视觉对齐约束。论文同时发布 GRIT-Bench，按文本长度和推理深度系统评估 text-rich 推理，实验显示 GRIT 在推理准确率和 grounding 保真度上均优于现有 RLVR 方法。
GRIT: Gradient Masking for Visually Misgrounded Tokens in Text-Rich Reasoning
Reinforcement learning with verifiable rewards (RLVR) has recently been extended to vision-language models (VLMs), yet its effectiveness remains limited for text-rich reasoning tasks. We highlight a key factor contributing to this limitation: models frequently arrive at correct final answers despite misperceiving visual information during intermediate steps. We define this phenomenon as Correct-but-Visually-Misgrounded Reasoning (CVMR). Because standard RLVR rewards only final-answer correctness, it fails to differentiate between properly grounded and misgrounded reasoning, reinforcing both equally. Through token-level analysis, we reveal that CVMR tokens exhibit distinct attention patterns that differ systematically from well-grounded ones. Based on this observation, we propose GRIT, a training framework that incorporates visual grounding into RLVR to mitigate the impact of CVMR. Furthermore, to address the lack of text-rich reasoning benchmarks that allow for systematic evaluation across text length and reasoning depth, we introduce GRIT-Bench. Experiments demonstrate that GRIT significantly improves both reasoning accuracy and grounding fidelity over existing RLVR methods.