TreeSoc: Tree-Structured Dynamic Reasoning and Tool Synergy for Soccer Video Understanding

2026-07-12

AI Summary

This work addresses the limitations of existing vision-language models in soccer video understanding—specifically, shallow cross-modal alignment, weak multi-step reasoning, and insufficient tool coordination—by proposing a tree-structured dynamic reasoning framework that reformulates complex video question answering as a hierarchical search task. Integrating dynamic depth-first search with modular tool ensembles, the method enables ordered subtask decomposition, iterative refinement of intermediate states, adaptive tool routing, and selective activation of domain-specific modules, substantially enhancing contextual awareness and cross-domain generalization. Evaluated on SoccerBench, the approach achieves accuracies of 85.2%, 87.4%, and 82.2% on TextQA, ImageQA, and VideoQA, respectively, and attains a cross-domain accuracy of 74.16% on NExT-QA.

Abstract

Automated understanding of complex soccer scenarios from video remains a significant challenge for contemporary vision-language models (VLMs), which suffer from shallow cross-modal alignment and exhibit fundamental limitations in multi-step reasoning and coordinated tool integration. We present TreeSoc, a structured reasoning framework that reformulates soccer video question answering as a hierarchical search problem rather than a single-pass prediction. Specifically, TreeSoc employs a dynamic depth-first search (DFS) mechanism that decomposes complex queries into sequentially ordered sub-tasks, enabling iterative reasoning refinement through explicit intermediate states. This tree-structured decomposition naturally supports adaptive tool routing, wherein domain-specific modules are selectively activated and their outputs incorporated at each reasoning node to produce contextually grounded predictions. On SoccerBench, TreeSoc achieves state-of-the-art performance, with accuracies of 85.2%, 87.4%, and 82.2% on TextQA, ImageQA, and VideoQA, respectively. Additionally, TreeSoc further demonstrates strong cross-domain generalization, attaining 74.16% accuracy on NExT-QA. These results establish structured, tool-augmented tree reasoning as an effective paradigm for robust video understanding. Code is available at: https://github.com/thanhnhan29/TreeSoc.

Problem

• soccer video understanding
• vision-language models
• multi-step reasoning
• tool integration
• cross-modal alignment

Innovation

• tree-structured reasoning
• dynamic DFS
• tool synergy
• adaptive tool routing
• hierarchical search