# PhysMent: An Interactive Approach For LLM Reasoning In Physics Problems

arXiv: https://arxiv.org/abs/2609.13152 · HTML: https://mlsanity.com/paper/2609.13152/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 6 of its 44 references that are in the map.
- First posted on arXiv 7 Jul 2026.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)

## Builds on (most cited first, 6 of 6)

- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [AI2-THOR: An Interactive 3D Environment for Visual AI](https://mlsanity.com/paper/1712.05474/) (2017)
- [BEHAVIOR: Benchmark for Everyday Household Activities in Virtual, Interactive, and Ecological Environments](https://mlsanity.com/paper/2108.03332/) (2021)
- [IntPhys: A Framework and Benchmark for Visual Intuitive Physics Reasoning](https://mlsanity.com/paper/1803.07616/) (2018)
- [Answer Matching Outperforms Multiple Choice for Language Model Evaluation](https://mlsanity.com/paper/2507.02856/) (2025)
- [ABench-Physics: Benchmarking Physical Reasoning in LLMs via High-Difficulty and Dynamic Physics Problems](https://mlsanity.com/paper/2507.04766/) (2025)

## Abstract (arXiv)

Large language models (LLMs) perform strongly on static science benchmarks, yet their ability to reason about the physical world through active experimentation remains poorly understood. We introduce PhysMent, a benchmark that evaluates LLM physical reasoning via iterative, toolmediated interaction with a MuJoCo physics simulator. Unlike static benchmarks that supply all quantities upfront, PhysMent requires models to discover information by applying forces, querying object states, advancing time, and modifying scene geometry before answering. The benchmark comprises 105 scenes of classical mechanics, organized across four difficulty regimes (Easy/Hard and Single/Multi), three scene modalities (standard, object creation, hidden objects), and a scene-manipulation category, evaluated with a six-dimensional scoring framework. Results show that current models perform reasonably well on qualitative single-concept tasks (up to 80% accuracy) but degrade substantially on quantitative tasks that demand precise, multi-step experimental procedures: most models fall below 30% on the hardest single-concept category, where the bottleneck is procedural (adaptive multi-step tool use) rather than conceptual load. Across the seven models, accuracy ranges from 25% to 67%, with failures due to premature answer submission, inefficient exploration, and inconsistent grounding in simulator feedback rather than conceptual gaps.