4B模型超越397B大模型！北大+中南联手破解多模态视觉理解难题
06-27 01:07
精选参考来源
精选参考来源
1. Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation通过 On-Po
Together_CZ 2026-05-25 00:00:00
2. （2026|北大 & 中南，以文本想象思考，MLLM，基于文本的想象，思维链，自蒸馏）Imagine-OPD：通过策略内自蒸馏实现视觉操作内化 原创
CSDN博客 2026-06-10 00:00:00
3. Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation - 道客巴巴
道客巴巴 2026-06-13 00:00:00
4. Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation通过 On-Po
Together_CZ 2026-05-31 00:00:00
5. Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation
arXiv 2026-05-27 00:00:00
6. Vision-OPD: 多模态精细细节OPD
小红书 2026-05-22 00:00:00
7. 鹅厂多模态OCR大模型技术演进与商业化应用概要
腾讯云 2026-05-30 00:00:00
8. 深度求索发布DeepSeek-OCR，突破新领域为LLM长上下文提供新思路
今日头条 2026-06-23 00:00:00
9. TPAMI 2026 | 视觉特征再升级！vMLLM：多级聚合 + 跨模态增强，MLLM 性能暴涨
新浪财经 2026-06-10 00:00:00
10. The Many Faces of On-Policy Distillation: Pitfalls, Mechanisms, and Fixes
arXiv 2026-05-11 00:00:00
11. 强老师救不了OPD 原创
CSDN博客 2026-04-21 00:00:00
12. Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
智源社区 2026-04-14 00:00:00
13. Thinking Without Images: Internalizing Visual Manipulation with On-Policy Self-Distillation
arXiv 2026-06-07 00:00:00
14. 小红书提出 Vision-OPD | 9B参数+6.2K数据，一次前向传播干翻GPT-5.4和Gemini-3.1-Pro
微信公众号 2026-05-21 00:00:00
15. 后训练新范式：为什么 DeepSeek-V4 放弃混合 RL，改用 OPD？
CSDN博客 2026-05-12 00:00:00
16. DiffusionOPD：复旦联合通义万相提出扩散模型在线策略蒸馏新范式
今日头条 2026-05-29 00:00:00
17. DiffusionOPD: A Unified Perspective of On-Policy Distillation in Diffusion Models
arXiv 2026-05-14 00:00:00
已收藏
去我的收藏夹