aiqbench analysis
This research addresses the persistent gap between LLMs' mathematical proficiency and their ability to perform genuine physics reasoning by introducing a framework that explicitly separates physical modeling from calculation. By distilling intermediate representations of physical systems and employing a two-stage post-training strategy involving supervised fine-tuning and reinforcement learning, the method significantly improves performance on multimodal benchmarks like PhysReason and PhyX. The findings suggest that explicit structural modeling is a highly efficient pathway for enhancing physics reasoning capabilities, particularly in smaller language models.
Decoupled Physical Modeling and Execution for Physics Reasoning
arXiv:2608.22126v1 Announce Type: cross Abstract: Physics reasoning requires constructing a consistent model of the underlying physical system rather than relying solely on symbolic or formula-based manipulation. Although large language models have shown strong ability in solving math and coding problems, they still struggle with physics problems, as these problems entangle the physical modeling process with mathematical calculations. Humans approach physics by first building a representation of the system before performing calculations. Inspired by this, we introduce a unified framework that distills intermediate representations that explicitly encode the physical modeling process and adopt a two-stage post-training strategy, where supervised fine-tuning establishes structured modeling, and reinforcement learning with rubric-based feedback improves the quality of the modeling process. Experiments on multiple multimodal physics benchmarks show that our approach leads to consistent improvements in reasoning performance across different models and datasets. On PhysReason, PhyX and SeePhys benchmarks, physical modeling output performs GRPO by an average ~3%, showing that explicit physical modeling is an efficient strategy of improving physics reasoning for small LLMs.
Abstract:Physics reasoning requires constructing a consistent model of the underlying physical system rather than relying solely on symbolic or formula-based manipulation. Although large language models have shown strong ability in solving math and coding problems, they still struggle with physics problems, as these problems entangle the physical modeling process with mathematical calculations. Humans approach physics by first building a representation of the system before performing calculations. Inspired by this, we introduce a unified framework that distills intermediate representations that explicitly encode the physical modeling process and adopt a two-stage post-training strategy, where supervised fine-tuning establishes structured modeling, and reinforcement learning with rubric-based feedback improves the quality of the modeling process. Experiments on multiple multimodal physics benchmarks show that our approach leads to consistent improvements in reasoning performance across different models and datasets. On PhysReason, PhyX and SeePhys benchmarks, physical modeling output performs GRPO by an average ~3%, showing that explicit physical modeling is an efficient strategy of improving physics reasoning for small LLMs.
Machine Learning (cs.LG); Computation and Language (cs.CL)
arXiv:2608.22126 [cs.LG]
(or arXiv:2608.22126v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.22126 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 261 words
Related coverage
The three ways AI unlocks transformation in Retail, Travel, and Consumer Goods
Databricks · Jul 20, 2026
Links #5: 2026/07
LessWrong · Jul 31, 2026
PhysCodeBench: Benchmarking Physics-Aware Symbolic Simulation of 3D Scenes via Self-Corrective Multi-Agent Refinement
arXiv · Sep 14, 2026
Think Like a Pilot: Fine-Grained Long-Horizon UAV Navigation
arXiv · Jun 8, 2026