aiqbench analysis
WebKD proposes a training method for smaller open-weight web agents that uses a closed teacher to edit the student's own failed episodes, rather than imitating offline demonstrations. The reported gains of roughly 5.5 to 10.7 points over behavior cloning, outcome-reward RL, and teacher-editing baselines are on TimeWarp, with some transfer to the live web, but these are single-paper results on specific benchmarks. Readers should watch whether the approach holds across other agent tasks and teacher models, and whether the cost savings from smaller students hold once editing and training costs are counted.
Teaching Agents in their Own Style: Blackbox Teacher Distillation for Web Agents
arXiv:2610.19834v1 Announce Type: new Abstract: Frontier web agents using closed models have high inference and computational costs. Smaller open models can be an alternative, but to be performant, they require training supervision from a human or teacher model. Token-level teacher distillation is not an option for closed teachers as it needs access to the model's logits. The standard approach is to imitate teacher demonstrations offline, but it trains the student on states it has never visited, and in a style it does not write in. We introduce WebKD, where the teacher is the editor of the student's failures. Given a failed episode, the teacher finds the first point of mistake, and rewrites the student's reasoning traces, harness states, and actions from that point onward. The editing reveals that the teacher's corrections can be more surprising to the student than the student's own text. WebKD instead aims to edit minimally. In the data space, the teacher writes each correction in the student's style. In the objective space, the reasoning is weighted by its contribution to the action. A role-asymmetric KL anchor allows the corrected action to move freely while keeping the reasoning near the student's prior. On the TimeWarp benchmark, WebKD students gain about 5.5-10.7 pp over behavior cloning, outcome-reward RL, and teacher editing baselines. The performance transfers to unseen versions and to the live web in the OnlineMind2Web benchmark. WebKD opens a new way of distilling black-box models to improve the capabilities of open-weight web agents.
Abstract:Frontier web agents using closed models have high inference and computational costs. Smaller open models can be an alternative, but to be performant, they require training supervision from a human or teacher model. Token-level teacher distillation is not an option for closed teachers as it needs access to the model's logits. The standard approach is to imitate teacher demonstrations offline, but it trains the student on states it has never visited, and in a style it does not write in. We introduce WebKD, where the teacher is the editor of the student's failures. Given a failed episode, the teacher finds the first point of mistake, and rewrites the student's reasoning traces, harness states, and actions from that point onward. The editing reveals that the teacher's corrections can be more surprising to the student than the student's own text. WebKD instead aims to edit minimally. In the data space, the teacher writes each correction in the student's style. In the objective space, the reasoning is weighted by its contribution to the action. A role-asymmetric KL anchor allows the corrected action to move freely while keeping the reasoning near the student's prior. On the TimeWarp benchmark, WebKD students gain about 5.5-10.7 pp over behavior cloning, outcome-reward RL, and teacher editing baselines. The performance transfers to unseen versions and to the live web in the OnlineMind2Web benchmark. WebKD opens a new way of distilling black-box models to improve the capabilities of open-weight web agents.
Artificial Intelligence (cs.AI)
arXiv:2610.19834 [cs.AI]
(or arXiv:2610.19834v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.19834 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 335 words
Related coverage
Projection sampling preserves capabilities but diminishes gains from alignment fine-tuning
LessWrong · Oct 16, 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · Oct 5, 2026
How can LLM RL Work Despite Information-Theoretic Inefficiency
LessWrong · Oct 16, 2026
The generative AI customization spectrum: From prompt engineering to custom models on AWS
AWS · Sep 14, 2026