BoT-Feedback: Grounding Multimodal Reasoning in Biomechanical Evidence for Explainable Human Action Feedback

2026-10-03

AI Summary

This study addresses the limitations of multimodal large language models in providing action feedback, specifically their tendency toward generic suggestions, poor interpretability, and physical hallucinations. To this end, we propose the BoT-Feedback framework, which introduces a novel four-stage reasoning paradigm termed "Biomechanics-of-Thought." This approach anchors multimodal reasoning within structured biomechanical evidence by integrating 3D skeleton extraction with temporal alignment techniques. Furthermore, the project develops a plug-and-play parser alongside the BiomAF benchmark dataset. Experimental results demonstrate that the proposed framework improves expert evaluation scores by 40% and significantly mitigates physical hallucinations. Notably, it enables smaller open-source models to achieve performance levels comparable to those of large proprietary systems.

Abstract

Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in visual understanding and multimodal reasoning, yet they remain fundamentally limited in Human Action Feedback Generation. Existing methods infer coaching feedback directly from visual observations, producing generic advice, limited interpretability, and physically implausible hallucinations. In contrast, expert human coaches diagnose performance through explicit biomechanical reasoning over joint kinematics, posture, and body dynamics. We introduce BoT-Feedback, a framework that grounds MLLM reasoning in structured biomechanical evidence. Our key contribution is Biomechanics of Thought (BoT), a four-stage reasoning framework that progressively identifies the action, localises the critical body regions, analyses quantitative biomechanical differences between expert and student performances, and synthesises interpretable coaching feedback. To support this reasoning process, we develop a plug-and-play Biomechanical Data Parser (BDP) that converts videos into structured biomechanical descriptors and an alignment strategy that temporally matches expert and student motions. We further introduce BiomAF, a benchmark containing paired teacher-student videos, 3D skeletons, biomechanical attributes, and expert-coaching annotations. Experiments across twelve open- and closed-source MLLMs demonstrate that grounding reasoning in biomechanical evidence consistently improves feedback quality, interpretability, and robustness while substantially reducing biomechanical hallucinations. BoT-Feedback improves the average expert evaluation score from 2.07 to 2.95 (+40%), enabling compact open-source MLLMs to approach the performance of substantially larger proprietary systems for explainable action feedback generation.

Problem

• Multimodal Large Language Models
• Human Action Feedback
• Biomechanical Hallucinations
• Explainability
• Visual Reasoning

Innovation

• Multimodal Large Language Models
• Biomechanics of Thought
• Human Action Feedback
• Explainable AI
• Biomechanical Data Parser