arXiv cs.AI· Alexander Yue, Qibin Liu, Julia Gonski·· 1 天前AI 评分34
用强化学习让智能体 AI 自我改进：面向对撞机物理的奖励驱动挑战
Reward-Driven Collider Physics Challenges for Self-Improving Agentic AI via Reinforcement Learning
AI 导读
研究提出一种开放式优化任务格式，用连续、自动评估的奖励把高能物理难题转化为强化学习环境，而非仅考察任务完成度的传统基准。作者在多项高能物理任务上评测了前沿推理模型，并给出初步验证：一个小语言模型无需人工干预，仅靠自动反馈的强化学习即可获得提升。目标是把智能体 AI 从复现既有工作流推进到自主发现更优科学解法。
正文 · 原文
Abstract:Recent advances in agentic AI have demonstrated the potential to automate complex high-energy physics (HEP) workflows. We focus on the regime of HEP tasks where frontier models still struggle to provide correct and complete solutions, and introduce an open-ended optimization task format with continuous, automatically evaluated rewards. Unlike conventional benchmarks focused on task completion, these tasks target reinforcement learning (RL) environments to generate agentic improvement on state- of-the-art HEP challenges. We benchmark frontier reasoning models on several HEP tasks and demonstrate an initial proof of concept in which a small language model improves through RL using automated feedback without human intervention. By connecting challenging HEP problems with verifiable scientific rewards and RL, we aim to advance agentic AI from reproducing established workflows to autonomously discovering better scientific solutions, paving the way toward self-improving scientific agents.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2611.15885 [cs.AI] |
| (or arXiv:2611.15885v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.15885 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Qibin Liu [view email]
[v1]
Wed, 11 Nov 2026 19:00:41 UTC (562 KB)
来源：arXiv cs.AI · arxiv.org