ScholarPulse 日报 2026-10-28

2026-10-28 学术简报：2 篇。评估显示LLM在无提示自由探索中发现物理规律的能力极弱，12个被测模型中仅GPT-6-Astra达到16.7%的发现率，其余均为0%。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	Can LLMs Uncover Hidden Physical Laws through Free Exploration?	arXiv	2026-10-27	AI-Agent	高
2	Agentic Distributional Regression: Conditional Density Estimation through Agent-Written Programs	arXiv	2026-10-27	RAG	高

重点论文与技术动态

1. Can LLMs Uncover Hidden Physical Laws through Free Exploration?

来源：arXiv
日期：2026-10-27
作者/机构：Zhen Yang, Wen Wang, Xander Xu, Ying-Cong Chen
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.CL, cs.AI, cs.MA

一句话结论

评估显示LLM在无提示自由探索中发现物理规律的能力极弱，12个被测模型中仅GPT-6-Astra达到16.7%的发现率，其余均为0%。

核心内容

提出LongChang自动化框架，在被修改物理法则的模拟环境中评估LLM的科学发现能力。
LLM作为探索者自主提问、交互并制造工具，配合其他评估智能体协同驱动场景演进，全程无需人工干预。
探索记录显示，模型普遍忽视未探究的物理异常，且仅能部分识别修改后的物理规律。

方法与数据

采用集成了场景生成、工具审查和实验报告评估的多智能体闭环框架；评测了12个LLM（除GPT-6-Astra外其他模型名称摘要未明确）。

价值判断

值得关注：开创了无需预设任务和提示、依托自由探索发现物理规律的全自动AI评测范式。
可复用点：由探索记录动态引导场景演化及报告验证的多智能体自动化机制。
局限/待核查：模型发现成功率极低，且难以主动深入调查环境中的反常物理现象。

摘要

Scientific discovery often begins with an unexpected observation that prompts a new research question. In a world with altered physical laws, can large language models (LLMs) notice unexpected behavior, such as heavier objects falling more slowly, and discover the law through experiments without being told what to investigate? To evaluate this ability, we introduce LongChang, an automated framework that creates simulated worlds with altered physics, where one or more laws differ from ordinary physics. The evaluated LLMs act as explorers, manipulating objects and building tools to investigate their own questions without research tasks or hints about these laws. Separate LLM agents generate and review scenes, review tool designs, and evaluate reports for correctness and experimental support using the altered physical laws and exploration records. These records guide scene evolution for further exploration, allowing the full cycle to run without human intervention. Among 12 LLMs, GPT-6-Astra achieves a discovery rate of 16.7%, while the other 11 score 0%. The exploration records reveal uninvestigated anomalies and only partial identification of the altered physical laws.

2. Agentic Distributional Regression: Conditional Density Estimation through Agent-Written Programs

来源：arXiv
日期：2026-10-27
作者/机构：Lorenzo Tedesco
主题标签：RAG, arXiv
推荐等级：高
分类：stat.ME, stat.ML

一句话结论

论文提出智能体分布回归（ADR），由大语言模型智能体编写简短程序生成显式公式来估计条件概率密度，理论完备且实证性能最优。

核心内容

智能体建模：智能体自主选择结构与拟合过程，编写将协变量映射为条件分布（参数分布、混合或多元分解）的简短程序。
公式可解释：受限于计算操作和常数数量，生成的条件密度为可脱离智能体、独立进行审查与计算的显式公式。
统计理论保证：给出了有限样本误差界、基于独立验证集选择的预言机不等式，以及稳定选择下的渐近正态性。

方法与数据

采用 Claude Opus 5.5 结合概率分布库编写拟合程序；在模拟实验、12个真实数据集及1个训练截止期后收集的新数据集上进行了评估。

价值判断

值得关注：开创了LLM自动编写公式化程序进行密度估计的范式，平均表现优于所考察的对比方法。
可复用点：以短程序表示显式参数化分布的设计，以及基于验证集筛选模型的统计分析框架。
局限/待核查：理论成立依赖于优化与稳定选择等假设；智能体运行开销与拟合耗时摘要未明确。

摘要

In this paper we propose a method for estimating conditional densities from an iid sample using a language-model agent. The agent writes the estimator as a short program that, for each covariate value, returns the estimated conditional distribution of the response. The program is built from a library of probability distributions, so the returned law can be a parametric distribution, a mixture or a multivariate factorization, with parameters given by explicit functions of the covariates. Since the number of operations and constants in the program is limited, the fitted conditional density is a formula that can be inspected and evaluated independently of the agent, while the choice of the model structure and of the fitting procedure is left to the agent. We call this approach agentic distributional regression (ADR). Treating the returned program as a random conditional density estimator, we derive a finite-sample error bound under explicit complexity and optimization assumptions, and an oracle inequality for selection using independent validation data. For a stably selected regular structure with sufficiently accurate fitting, we obtain asymptotic normality of the fitted constants. In simulation studies, on twelve real datasets and on a new dataset collected after the agents' training data, ADR with Claude Opus 5.5 has the best average performance among the methods considered.