arXiv cs.AI· Atharva Naik, Vivek Chari, Pol Llado, Daniel Fried, Carolyn Ros\'e, Benjamin Van Durme, Sam Thomson, Justin Svegliato·· 12 小时前AI 评分34
AdaPRA：面向数据科学、带自适应评分标准的流程奖励智能体
AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science
AI 导读
AdaPRA 是一个持久化编码智能体框架，通过跨尝试积累知识并用自适应评分标准做流程监督，审计智能体轨迹与工作区以识别和纠正过程错误。每任务五次尝试下，它让 80B Qwen 模型在 DSPredict-Easy 排行榜百分位提升 24.5 点、DSPredict-Hard 提升 14.8 点，同时减少过程错误。
正文
Abstract:Data science agents can automate end-to-end machine learning workflows, yet they remain far from competitive on human leaderboards. Agents make process errors such as insufficient exploration, unsound validation, coding bugs, and poor use of a fixed budget (re-running the same exploratory data analysis and baseline models in every attempt). We propose AdaPRA to address these issues: a persistent coding agent harness that accumulates knowledge across attempts on the same task, with rubric-guided process supervision that identifies and corrects errors. Our process reward agent audits the agent's trajectory and workspace by executing checks, and adapts its rubrics to each task. We find that rubrics seeded from easier tasks generalize to harder tasks, with rubric adaptation further improving performance. With five attempts per task, AdaPRA improves an 80B Qwen model's leaderboard percentile by 24.5 points on DSPredict-Easy and 14.8 points on DSPredict-Hard while reducing process errors. We also show that supervision with rubrics mined from one base model's runs improves another (DeepSeek-V3.2), and that AdaPRA beats DataPRM, a competitive PRM for data science. We also share our code for reproducibility.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.33245 [cs.AI] |
| (or arXiv:2610.33245v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.33245 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Atharva Naik [view email]
[v1]
Fri, 23 Oct 2026 19:46:55 UTC (1,508 KB)
来源：arXiv cs.AI · arxiv.org