视觉对比自蒸馏

摘要

基于策略的自蒸馏方法（OPSD）因无需外部教师模型而优于基于策略的蒸馏方法（OPD），但其仍需师生模型间的信息不对称性，以确保自教师提供的学习信号强于学生模型。现有方法通过特权答案或视觉证据构建这种不对称性。我们探讨是否可同时消除这两者，从而实现一种更简洁的、纯粹由输入条件驱动的OPSD形式。为此，我们提出视觉对比自蒸馏方法（VCSD），其将图像内容去除转化为基于策略的自蒸馏信号。在学生模型生成的每个响应前缀处，指数移动平均（EMA）教师模型在同一提示和前缀下生成两种下一个词的分布——一种以原始图像为条件，另一种以内容擦除后的控制图像为条件。两种分布的逐词对数概率差凸显了因实例级视觉内容而可能性显著提升的候选词。我们利用这种对比，在教师模型原始图像分布的合理支持域内对其进行锐化处理，并将所得完整分布目标蒸馏至学生模型。使用ViRL39K数据集，VCSD在Qwen3-VL和Qwen3.5模型上的一致表现均优于匹配的OPSD方法。例如，在Qwen3-VL上，VCSD将七个基准测试的综合得分分别从2B参数的62.27%提升至67.04%、4B参数的71.30%提升至73.16%、8B参数的72.51%提升至76.26%。此外，VCSD无需外部教师模型、特权答案、视觉证据信号、推理轨迹，也不增加额外的推理时间成本。

English

On-policy self-distillation (OPSD) is promising as it removes the external teacher required by on-policy distillation (OPD), yet it still needs asymmetric information between teacher and student to ensure that the self-teacher provides a stronger learning signal than the student. Existing methods create this asymmetry either through privileged answers or visual evidence. We ask whether both can be removed, yielding a simpler form of OPSD driven purely by input conditioning. For this purpose, we propose Visual Contrastive Self-Distillation, namely VCSD, which converts image-content removal into an on-policy self-distillation signal. At each student-generated response prefix, the EMA teacher produces two next-token distributions under the same prompt and prefix -- one conditioned on the original image and the other on a content-erased control. Their token-wise log-probability difference highlights candidates whose likelihood is specifically increased by the instance-level visual content. We use this contrast to sharpen the teacher's original-image distribution within its plausible support, and distill the resulting full-distribution target into the student. Using ViRL39K dataset, VCSD consistently outperforms matched OPSD across Qwen3-VL and Qwen3.5 models. For example, on Qwen3-VL, it improves the seven-benchmark aggregate from 62.27% rightarrow 67.04% at 2B, 71.30% rightarrow 73.16% at 4B, and 72.51% rightarrow 76.26% at 8B. Furthermore, VCSD requires no external teacher, privileged answers, visual evidence signals, reasoning traces, or additional inference-time cost.