arXiv AI

Discovering High Level Patterns from Simulation Traces

The paper proposes an unsupervised learning approach that uses program synthesis to translate detailed simulation traces into sparse, high‑level structural patterns, making them easier for large language models to interpret. These pattern detectors can be guided by human‑provided labels such as "rigid collision" or "stretching spring" and produce transparent, explainable functions mapping system states to concise annotations. Experiments on a physics benchmark show that the annotated representations improve natural language reasoning about specific physical systems and enable natural‑language goals to be converted into reward programs for solution search.

arXiv Computation and Language

Aug 25

Decoupled Physical Modeling and Execution for Physics Reasoning

The paper introduces a framework that separates physical modeling from execution in physics reasoning tasks. It uses a two‑stage post‑training approach: supervised fine‑tuning to build structured models and reinforcement learning with rubric‑based feedback to refine them. Experiments on PhysReason, PhyX, and SeePhys show that this explicit modeling improves reasoning performance by about 3% on average for small LLMs.

By Ye Zhang, Xuehang Guo, Rui Pan, Pengfei Yu, Denghui Zhang, Manling Li, Qingyun Wang

llms reinforcement-learning robotics fine-tuning multimodal benchmarks

arXiv Machine Learning

Aug 14

Training and Benchmarking Code Generation for Physics-Inspired Animations

arXiv:2602. 10840v2 Announce Type: replace Abstract: Large language models (LLMs) have been widely studied in areas such as mathematical reasoning, complex coding, and scientific problem solving.

By Yanan Wang, Renxi Wang, Yongxin Wang, Xuezhi Liang, Fajri Koto, Timothy Baldwin, Xiaodan Liang, Haonan Li

llms reinforcement-learning multimodal benchmarks data-engineering

arXiv AI

Jun 2

Towards a Physics Foundation Model

arXiv:2509. 13805v4 Announce Type: replace-cross Abstract: Foundation models have revolutionized natural language processing through a ``train once, deploy anywhere'' paradigm, where a single pre-trained model adapts to countless downstream tasks without retraining.

By Florian Wiesner, Zo\"e J. Gray, Matthias Wessling, Stephen Baek

llms nlp

arXiv AI

Aug 5

Uncovering Spontaneous Physics Representations in In-Context Learning

arXiv:2508. 12448v2 Announce Type: replace-cross Abstract: In-context learning (ICL) lets large language models (LLMs) solve new tasks from prompts alone, across an ever-widening range of domains, yet the mechanisms underlying this ability remain poorly understood.

By Yeongwoo Song, Jaeyong Bae, Dong-Kyum Kim, Hawoong Jeong

llms

arXiv Machine Learning

Aug 31

Aligning Agentic World Models via Knowledgeable Experience Learning

arXiv:2601.13247v2 Announce Type: replace-cross Abstract: Current Large Language Models (LLMs) exhibit a critical modal disconnect: they possess vast semantic knowledge but lack the procedural ground...

By Baochang Ren, Yunzhi Yao, Rui Sun, Shuofei Qiao, Ningyu Zhang, Huajun Chen

llms agents fine-tuning safety

arXiv Computer Vision

Aug 31

Code as Worlds: Agentic Discovery of Executable World Representations for Physical Reasoning

arXiv:2608.27549v1 Announce Type: new Abstract: Physical understanding and reasoning depend on forming compact and generalizable representations of the world. While modern vision-language models can...

By Hanyang Wang, Yimo Cai, Weiliang Chen, Jiawei Chi, Haowen Sun, Qiyu Dai, Yi-Hsin Hung, Xingzhuo Guo, Jinshan Ren, Runmao Yao, Ziwei Liu, Mingsheng Long, Yueqi Duan, Jun Gao, Jiangran Lyu, Fangfu Liu, Jialong Wu

llms agents multimodal benchmarks

arXiv AI

Jul 28

Loong: Synthesize Long Chain-of-Thoughts at Scale through Verifiers

arXiv:2509. 03059v2 Announce Type: replace-cross Abstract: Recent advances in Large Language Models (LLMs) have shown that their reasoning capabilities can be significantly improved through Reinforcement Learning with Verifiable Reward (RLVR), particularly in domains like mathematics and programming, where ground-truth correctness can be automatically evaluated.

By Xingyue Huang, Rishabh, Gregor Franke, Ziyi Yang, Jiamu Bai, Weijie Bai, Jinhe Bi, Zifeng Ding, Yiqun Duan, Chengyu Fan, Wendong Fan, Xin Gao, Ruohao Guo, Yuan He, Zhuangzhuang He, Xianglong Hu, Neil Johnson, Bowen Li, Fangru Lin, Siyu Lin, Tong Liu, Yunpu Ma, Hao Shen, Hao Sun, Beibei Wang, Fangyijie Wang, Hao Wang, Haoran Wang, Yang Wang, Yifeng Wang, Zhaowei Wang, Ziyang Wang, Yifan Wu, Zikai Xiao, Chengxing Xie, Fan Yang, Junxiao Yang, Qianshuo Ye, Ziyu Ye, Guangtao Zeng, Yuwen Ebony Zhang, Zeyu Zhang, Zihao Zhu, Bernard Ghanem, Philip Torr, Guohao Li

llms agents reinforcement-learning benchmarks

arXiv AI

Jun 4

Simulate, Reason, Decide: Scientific Reasoning with LLMs for Simulation-Driven Decision Making

arXiv:2606. 04505v1 Announce Type: new Abstract: Scientific simulators are increasingly being integrated into LLM-driven systems for high-stakes simulation-driven decision-making.

By Yuhan Yang, Ruipu Li, Alexander Rodr\'iguez

llms agents

arXiv AI

Jul 7

Formal Disco: Scalable Open-Ended Generation of Formally Verified Programs

arXiv:2607. 04631v1 Announce Type: new Abstract: The cost of producing code is rapidly diminishing with increasingly capable AI agents, while quality assurance of generated programs has not kept pace.

By Gabriel Poesia, Simon Henniger, Tzu-Han Hsu, Yilun Du, Nada Amin

llms agents fine-tuning efficiency

arXiv AI

Aug 5

Large language models for partial differential equation workflows

arXiv:2608. 03600v1 Announce Type: new Abstract: Partial differential equations (PDEs) become actionable in science and engineering not as isolated formulae, but as executable workflows that connect modelling assumptions, governing equations, numerical solvers, diagnostics, and decisions.

By Han Wan, Rui Zhang, Hao Sun

llms benchmarks

arXiv Machine Learning

Sep 2

DiscoverPhysics: Benchmarking LLMs for Out-of-the-Box Scientific Thinking

arXiv:2605.26087v2 Announce Type: replace-cross Abstract: Frontier LLMs now perform strongly across a wide range of physics evaluations, but it is hard to disentangle genuine reasoning from recall of...

By Matt L. Wiemann, Lindsay M. Smith, Peter Melchior, Siddharth Mishra-Sharma, Andrew Gordon Wilson, Pavel Izmailov, Carolina Cuesta-L\'azaro

llms agents benchmarks

arXiv AI

Sep 16

The Imitation Game: When LLMs Learn to Reason Like Programs via Code-Centric Reasoning Data Synthesis

The paper introduces MIMIC, a framework that uses executable code to generate rigorous reasoning data for large language models (LLMs). By converting algorithms into verifiable reasoning trajectories through narrative fusion, code-guided test synthesis, and dynamic code instrumentation, MIMIC creates a Code-Instrumented Reward (CIR) that supplies dense, high‑fidelity supervision for reinforcement learning. Models trained with MIMIC’s synthetic dataset show significant, consistent improvements in general reasoning, complex mathematics, and fine‑grained deterministic tasks.

By Jinyang Zhang, Weibin Liao, Keqin Bao, Sihang Li, Shaobo Wang, Muyang Ye, Hongxin Ding, Yue Fang, Tianyi Tang, Fei Huang, Kexin Yang, Xingzhang Ren, Dayiheng Liu

llms reinforcement-learning fine-tuning benchmarks

Related stories