arXiv cs.CL· Marius Memmel, Shawn Yin, Urvi Rutia, Angad Wadhwa, Eldon Li, Gavin Angeloff, Ankit Goyal, Jesse Zhang, Dieter Fox, Fabio Ramos, Anqi Li, Abhishek Gupta·· 11 小时前AI 评分38
D&A：解耦描述与评估，让 VLM 成为零样本机器人奖励模型
D&A: Decoupling Description and Assessment Enables VLMs as Zero-Shot Robotic Reward Models
AI 导读
D&A 将机器人任务进度估计拆成两步：先由 VLM 生成客观的场景动态文本描述，再用显式评分标准做评估，从而把判断锚定在外部绝对成功定义上，而非视频内部时间线。该纯提示词方法校准了零样本估计并缓解确认偏差，能给出细粒度部分进度、准确惩罚失败执行，跨任务、跨本体与跨基准匹配或超越经后训练的奖励模型，其奖励估计还可支撑下游强化学习。
正文 · 原文
Authors:Marius Memmel, Shawn Yin, Urvi Rutia, Angad Wadhwa, Eldon Li, Gavin Angeloff, Ankit Goyal, Jesse Zhang, Dieter Fox, Fabio Ramos, Anqi Li, Abhishek Gupta
Abstract:Vision-Language Models (VLMs) hold promise for zero-shot video task-progress estimation in robot learning, yet existing frameworks frequently suffer from confirmation bias, hallucinating progress in failed executions due to implicit success assumptions or relative temporal ranking designs. Naive prompting of frontier models collapses into a binary success detector that fails to capture nuanced, non-linear progress and failures. While post-training approaches can mitigate this through explicit training, they demand expensive real-world robotics data and require computationally intensive post-training for every individual base model. To resolve these limitations, we introduce D&A (Describe & Assess): our approach decouples progress estimation into an objective, task-grounded textual description of scene dynamics followed by an assessment phase using explicit rubrics. By compelling the VLM to synthesize an explicit, text-based definition of absolute success before scoring, the framework anchors evaluations to an external, global criterion rather than the video's internal timeline. Empirical results demonstrate that our method calibrates zero-shot estimates and mitigates confirmation bias, providing fine-grained, partial progress metrics that accurately penalize failed executions while correctly interpolating successful runs. Being purely prompt-based, D&A generalizes across tasks, embodiments, and benchmarks---matching or surpassing post-trained reward models---and its reward estimates enable efficient downstream reinforcement learning.
| Comments: | Website at this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Systems and Control (eess.SY) |
| Cite as: | arXiv:2611.03622 [cs.RO] |
| (or arXiv:2611.03622v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2611.03622 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Marius Memmel [view email]
[v1]
Mon, 2 Nov 2026 18:52:11 UTC (23,156 KB)
来源：arXiv cs.CL · arxiv.org