Learning the Task or Learning the Verifier? Rethinking Criteria Exposure in Embodied Self-Improvement
Learning the Task or Learning the Verifier? Rethinking Criteria Exposure in Embodied Self-Improvement
Exposing benchmark success criteria to self-improving embedding-encoded intelligent agents can make them achieve higher success rates under the criteria, but does not improve the completion of instruction tasks, and the advantage weakens during standard fine-tuning or cross-task, while standards that are loyal to instructions better reflect task performance.
Published on Nov 5
Authors:
,
,
,
,
,
,
,
,
,
,
Abstract
Exposing benchmark success criteria to self-improving embodied coding agents can raise measured success under those criteria without improving instructed-task completion, and the advantage weakens under slight criteria changes and transfers poorly across tasks, while instruction-faithful criteria better reflect instructed-task performance.
Self-improving embodied agents are increasingly being developed and compared on robot simulation benchmarks. Several recent works expose the benchmark's success criteria to the learning agent, a practice we call Criteria Exposure (CE). These criteria, originally written to verify policies trained by imitation learning, often demand more than the corresponding language instructions, so a score from such a verifier may not faithfully reflect whether the instructed task was completed, especially for robot-control programs written by an agent. As embodied coding agents report ever higher benchmark scores, a natural question arises: when a coding agent is exposed to the benchmark's criteria, does it generate code that better performs the instructed task, or code that simply overfits to the verification criteria? We conduct controlled experiments with a self-improving agent with and without CE, and find that with CE the agent quickly produces code with a higher success rate under the given criteria, but this advantage disappears when the criteria are changed slightly without departing from the original instructions. More importantly, the skills the agent develops under CE transfer poorly across tasks. We further propose instruction-faithful criteria and show that they better reflect the performance of agent-generated programs.
View arXiv page View PDF Project page Add to collection
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2611.08318 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2611.08318 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2611.08318 in a Space README.md to link it from this page.
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
Source: Hugging Face Daily Papers · huggingface.co