Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning

Haozhi Yuan, Handong Li, Yifan Zhang, Qianshan Wei

Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning: 40 upvotes on Hugging Face Daily Papers, #4 of 39 papers on 2026-11-10. Day-by-day upvote history.

Upvotes over time

It had 37 upvotes by its first day, 39 upvotes by the next day, 40 upvotes by day 3, 40 upvotes by day 7, and has 40 now. It ranked #14 of 133 papers in its week and #43 of 340 in its month. Its code repository has 34 GitHub stars.

Agentic reinforcement learning (RL) and multi-teacher on-policy distillation (MOPD) offer complementary routes to post-training. For visual agents, however, tool use presents two related challenges. Models may overlook useful tool-returned observations, while poorly grounded actions can limit the evidence available for subsequent distillation. We introduce Reframe, a post-training framework that treats the observation state, rather than only the output distribution, as an object of alignment. First, Ground-Drop-GRPO (GD-GRPO) trains spatial and temporal specialists with evidence-aware rewards and stochastic removal of the original visual input, encouraging them to acquire grounded evidence and use the returned observations. Second, Observation-Gated-MOPD (OG-MOPD) replaces deficient tool actions with canonical ones mid-rollout and resumes distillation from the repaired observation. The two stages complement each other, as learning to rely on tool-returned observations can help the policy benefit from observation repair. Neither stage requires a task-specific SFT cold start. Extensive experiments over 14 image and video benchmarks demonstrate that Reframe outperforms both Mix-RL and vanilla MOPD. Ablation results show that GD-GRPO increases observation dependence and tool gain compared with outcome-based GRPO.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.