MA-Bench: Towards Fine-grained Micro-Action Understanding

2026-03-27

AI Summary

This work addresses the absence of dedicated evaluation benchmarks for fine-grained micro-action understanding in current multimodal large language models (MLLMs), which hinders the assessment of their perceptual and reasoning capabilities regarding subtle human behaviors. To this end, we introduce MA-Bench—the first fine-grained benchmark specifically designed for micro-action understanding—comprising 1,000 videos and 12,000 structured question-answer pairs, along with a complementary training set, MA-Bench-Train, containing 20.5K videos. A three-tiered evaluation framework systematically measures model performance across micro-action perception, relational understanding, and explanatory reasoning. Experiments on 23 prominent MLLMs reveal significant limitations in modeling fine-grained action dynamics and body-part interactions. Notably, fine-tuning Qwen3-VL-8B on MA-Bench-Train substantially improves performance on reasoning and explanation tasks, demonstrating the benchmark’s effectiveness and practical utility.

Abstract

With the rapid development of Multimodal Large Language Models (MLLMs), their potential in Micro-Action understanding, a vital role in human emotion analysis, remains unexplored due to the absence of specialized benchmarks. To tackle this issue, we present MA-Bench, a benchmark comprising 1,000 videos and a three-tier evaluation architecture that progressively examines micro-action perception, relational comprehension, and interpretive reasoning. MA-Bench contains 12,000 structured question-answer pairs, enabling systematic assessment of both recognition accuracy and action interpretation. The results of 23 representative MLLMs reveal that there are significant challenges in capturing motion granularity and fine-grained body-part dynamics. To address these challenges, we further construct MA-Bench-Train, a large-scale training corpus with 20.5K videos annotated with structured micro-action captions for fine-tuning MLLMs. The results of Qwen3-VL-8B fine-tuned on MA-Bench-Train show clear performance improvements across micro-action reasoning and explanation tasks. Our work aims to establish a foundation benchmark for advancing MLLMs in understanding subtle micro-action and human-related behaviors. Project Page: https://MA-Bench.github.io

Problem

• Micro-Action Understanding
• Multimodal Large Language Models
• Benchmark
• Fine-grained Action Recognition
• Human Behavior Analysis

Innovation

• Micro-Action Understanding
• Multimodal Large Language Models
• Fine-grained Benchmark
• Structured Captioning
• Interpretive Reasoning