# Visual Contrastive Self-Distillation

Page: https://stenobird.com/podcast/daily-paper-cast-7079649/visual-contrastive-self-distillation
Text version: https://stenobird.com/podcast/daily-paper-cast-7079649/visual-contrastive-self-distillation.md
Podcast: [Daily Paper Cast](https://stenobird.com/podcast/daily-paper-cast-7079649)
Published: 2026-07-25T03:25:26+00:00
Episode link: https://share.transistor.fm/s/36304bfd
Audio file: https://media.transistor.fm/36304bfd/bfdbab39.mp3
Processing state: not_requested
JSON: https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/visual-contrastive-self-distillation
Duration seconds: 1284

## Resource

🤗 Upvotes: 41 | cs.CV, cs.AI Authors: Yijun Liang, Yunjie Tian, Yijiang Li, Yuqi Jia, Furong Huang, Tianyi Zhou, Di Fu Title: Visual Contrastive Self-Distillation Arxiv: http://arxiv.org/abs/2607.21556v1 Abstract: 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 additio…

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/visual-contrastive-self-distillation/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/daily-paper-cast-7079649/visual-contrastive-self-distillation.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.