Generate, Analyze, and Refine: Training-Free Sound Source Localization via MLLM Meta-Reasoning
On the map: LLM reasoning → Audio-language model reasoning
Abstract (arXiv)
Sound source localization task aims to identify the locations of sound-emitting objects by leveraging correlations between audio and visual modalities. Most existing SSL methods rely on contrastive learning-based feature matching, but lack explicit reasoning and verification, limiting their effectiveness in complex acoustic scenes. Inspired by human meta-cognitive processes, we propose a training-free SSL framework that exploits the intrinsic reasoning capabilities of Multimodal Large Language Models (MLLMs). Our Generation-Analysis-Refinement (GAR) pipeline consists of three stages: Generation produces initial bounding boxes and audio classifications; Analysis quantifies Audio-Visual Consistency via open-set role tagging and anchor voting; and Refinement applies adaptive gating to prevent unnecessary adjustments. Extensive experiments on single-source and multi-source benchmarks demonstrate competitive performance. The source code is available at https://github.com/VisualAIKHU/GAR-SSL.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Audio-language model reasoning, one of 17 directions in LLM reasoning. Builds on 44 of its 51 references that are in the map. First posted on arXiv 8 Apr 2026.
Builds on
44 of 51 references.
Closest ancestors · 1
Same direction, newest first
- 2025Qwen2.5-Omni Technical Report26 Mar
Other references · 43
- 2015Deep Residual Learning for Image Recognitioncited by 52,096
- 2024GPT-4o System Cardcited by 4,588
- 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,386
- 2023A Survey on Multimodal Large Language Modelscited by 909
- 2023Multimodal Chain-of-Thought Reasoning in Language Modelscited by 752
- 2020VGGSound: A Large-scale Audio-Visual Datasetcited by 648
- 2018Audio-Visual Scene Analysis with Self-Supervised Multisensory Featurescited by 335
- 2018The Sound of Pixelscited by 259
- 2017Objects that Soundcited by 256
- 2022MM-Diffusion: Learning Multi-Modal Diffusion Models for Joint Audio and Video Generationcited by 227
- 2023Multimodal Large Language Models: A Surveycited by 195
- 2018Learning to Localize Sound Source in Visual Scenescited by 189
31 more
- 2025MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipecited by 186
- 2021Localizing Visual Sounds the Hard Waycited by 170
- 2020Multiple Sound Sources Localization from Coarse to Finecited by 118
- 2020Discriminative Sounding Objects Localization via Self-supervised Audiovisual Matchingcited by 104
- 2022Localizing Visual Sounds the Easy Waycited by 82
- 2022A Closer Look at Weakly-Supervised Audio-Visual Source Localizationcited by 73
- 2025LLaVA-ST: A Multimodal Large Language Model for Fine-Grained Spatial-Temporal Understandingcited by 59
- 2023Large Language Models are Visual Reasoning Coordinatorscited by 58
- 2022Mix and Localize: Localizing Sound Sources in Mixturescited by 54
- 2023Audio-Visual Grouping Network for Sound Localization from Mixturescited by 52
- 2023Learning Audio-Visual Source Localization via False Negative Aware Contrastive Learningcited by 47
- 2022Exploiting Transformation Invariance and Equivariance for Self-supervised Sound Localisationcited by 44
- 2022Learning Sound Localization Better From Semantically Similar Samplescited by 32
- 2022Self-Supervised Predictive Learning: A Negative-Free Method for Sound Source Localization in Visual Scenescited by 31
- 2025InteractiveOmni: A Unified Omni-modal Model for Audio-Visual Multi-turn Dialoguecited by 29
- 2021Unsupervised Sound Localization via Iterative Contrastive Learningcited by 28
- 2024Towards Open-Vocabulary Audio-Visual Event Localizationcited by 26
- 2024T-VSL: Text-Guided Visual Sound Source Localization in Mixturescited by 19
- 2022Hear The Flow: Optical Flow-Based Self-Supervised Visual Sound Source Localizationcited by 14
- 2024Learning to Visually Localize Sound Sources from Mixtures without Prior Source Knowledgecited by 14
- 2025MMReason: An Open-Ended Multi-Modal Multi-Step Reasoning Benchmark for MLLMs Toward AGIcited by 14
- 2025Corvid: Improving Multimodal Large Language Models Towards Chain-of-Thought Reasoningcited by 11
- 2025Human-centered Interactive Learning via MLLMs for Text-to-Image Person Re-identificationcited by 9
- 2025ReasoningTrack: Chain-of-Thought Reasoning for Long-term Vision-Language Trackingcited by 9
- 2024Sim2Real Transfer for Audio-Visual Navigation with Frequency-Adaptive Acoustic Field Predictioncited by 7
- 2023Audio-Visual Spatial Integration and Recursive Attention for Robust Sound Source Localizationcited by 5
- 2025Object-aware Sound Source Localization via Audio-Visual Scene Understandingcited by 5
- 2024MLLM-Search: A Zero-Shot Approach to Finding People using Multimodal Large Language Modelscited by 4
- 2023Zero- and Few-shot Sound Event Localization and Detectioncited by 3
- 2025OrderChain: Towards General Instruct-Tuning for Stimulating the Ordinal Understanding Ability of MLLMcited by 3
- 2025Sound Source Localization for Human-Robot Interaction in Outdoor Environmentscited by 2
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Hearing and Seeing Through CLIP: A Framework for Self-Supervised Sound Source Localization
- 2026The Sonar Moment: Benchmarking Audio-Language Models in Audio Geo-Localization
- 2026What Did the MLLM Hear? Token-Level Spectro-Temporal Grounding for Audio MLLM Explainability
- 2026The World is Not Mono: Enabling Spatial Understanding in Large Audio-Language Models
- 2025Thinking with Sound: Audio Chain-of-Thought Enables Multimodal Reasoning in Large Audio-Language Models
- 2024BAT: Learning to Reason about Spatial Sounds with Large Language Models
- 2026Interpretable Perception and Reasoning for Audiovisual Geolocation
- 2026SonicWorld: Source-Centric Latent Dynamics for Spatial Audio Reasoning