大模型在脑子里偷偷推理，现在终于有人盯着它看了
大模型做视觉推理时，很多步骤是在“脑子里”用连续向量完成的，不写成文字。这带来一个麻烦：没人看得见它到底在算什么，也就没法监督它有没有真的在看图。研究者发现一个“证据信用缺口”:当你把图片里决定答案的关键细节改掉，模型那些隐藏推理向量几乎没反应——说明它根本没在认真看，只是蒙对了答案。他们提出一套方法，用“正确答案 vs 模型自己答错的答案”和“相关图片区域 vs 不相关区域”做对照，把监督信号直接灌进隐藏推理轨迹里。在三个模型家族上，这套方法都超过现有方案，并且第一次把这种“隐式推理”做到了235B参数的顶级模型规模，效果依然稳定。它不是你明天能直接用的东西，但它指向一个关键趋势：AI的推理正在从“说给你听”转向“自己算”，而如何让这种看不见的推理可靠，会成为接下来大模型竞争的核心问题。
📄 原文摘要(英文)
Latent visual reasoning (LVR) enables multimodal large language models (MLLMs) to perform intermediate computation in continuous latent tokens rather than expressing every reasoning step in words. However, unlike textual CoT, latent reasoning is not directly observable, making it difficult to supervise what latent tokens learn. In this work, we first conduct a thorough analysis of latent-token behavior and identify a latent evidence-credit gap: latent tokens respond only weakly to image perturbations that alter the correct answer. We hypothesize that this issue stems from the lack of explicit supervision during GRPO training. These findings suggest that a final-answer reward provides too little guidance on what visual evidence to preserve or how credit should be assigned across latent tokens. To bridge this gap, we propose ReaLVR, which brings visual-evidence supervision to the model's own free-running latent trajectories. ReaLVR contrasts correct and model-generated wrong answers to determine where stronger supervision is needed, and relevant and mismatched visual evidence to specify what to preserve. Across three model families, ReaLVR consistently outperforms evaluated LVR baselines, achieving the highest five-task average of 63.7% on Qwen2.5-VL-7B. Crucially, we are the first to scale visual reasoning in latent space, showing that our framework continues to deliver robust improvements at frontier model scales up to 235B. Further analyses show more question-sensitive latent-token positions, stronger alignment with relevant visual regions, and greater fixed-context dependence on the most attended latent tokens.