Search papers, labs, and topics across Lattice.
This paper introduces AdaptRubric, a Coarse-to-Fine Rubrics Framework designed to enhance GUI reward modeling by creating task-adaptive judging criteria. By implementing a two-stage process that first retrieves category-level rubrics and then generates instance-specific criteria, AdaptRubric addresses the limitations of existing methods that fail to adequately adapt to individual task requirements. The results demonstrate a significant performance improvement, with a 3.6-point increase in F1 score and a 4.23-point gain in task success compared to baseline models.
AdaptRubric's innovative two-stage framework boosts GUI reward modeling performance by over 3.6 F1 points, showcasing the power of task-adaptive criteria.
Recent studies on GUI agents have increasingly focused on outcome reward modeling, which assigns outcome rewards by judging whether an executed trajectory satisfies the success criteria implied by the user instruction. Existing GUI reward verifiers, however, often under-specify how these criteria should be constructed for each task instance. Whether using generic rubric structures or implicit model reasoning, their judging criteria are not sufficiently task-adaptive: they can transfer checks across tasks, overlook concrete constraints in the current instruction, or become overly strict by enforcing unstated requirements. To address this limitation, we propose AdaptRubric, a Coarse-to-Fine Rubrics Framework that constructs task-adaptive judging criteria through a category-level coarse stage and an instance-level fine stage. AdaptRubric performs category-level coarse rubric retrieval by routing the instruction to a GUI task family and retrieving reusable task-family criteria, then conducts instance-level fine rubric generation to surface compact cues for concrete values, scopes, and constraints in the current instruction. Across offline reward evaluation and online reinforcement learning optimization, AdaptRubric consistently outperforms prior reward agents, improving F1 by 3.6 points over the baseline average under a matched image budget and yielding a 4.23-point task-success gain.
