Does Location Matter? A Black-Box Causal Audit Reveals Model-Dependent Audio Grounding in LALM Hallucination

arXiv:2610.13112 · cs.AI, cs.LG, cs.SD, eess.AS · Submitted 2026-10-09 · Read on arXiv

cs.AI, cs.LG, cs.SD, eess.AS

Submitted: 2026-10-09

Updated: 2026-10-09

Comments: 5 pages, 1 figures, 4 tables. Submitted to IEEE ICASSP 2027

Code: https://github.com/khalooei/causa

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Grounding — 1× in this paper · explained in 3 episodes
• Hallucination — 1× in this paper · explained in 4 episodes

Sources

• Reducing Object Hallucination in Large Audio-Language Models via Audio-Aware Decoding
• Qwen2-Audio Technical Report
• SALMONN: Towards Generic Hearing Abilities for Large Language Models
• Understanding Sounds, Missing the Questions: The Challenge of Object Hallucination in Large Audio-Language Models
• Can Large Audio-Language Models Truly Hear? Tackling Hallucinations with Multi-Task Assessment and Stepwise Audio Reasoning
• COPO: Causal-Oriented Policy Optimization for Hallucinations of MLLMs
• Causal Tracing of Audio-Text Fusion in Large Audio Language Models
• Tracing Audio Grounding and Answer Selection in Audio LLMs
• Encode Once, Decode Never: Reusing Audio LM Internals for Efficient Temporal Localization
• Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models
• Qwen2.5-Omni Technical Report
• Calibrate Before Use: Improving Few-Shot Performance of Language Models
• Multi-Modal Hallucination Control by Visual Information Grounding
• Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding
• Temporal Contrastive Decoding: A Training-Free Method for Large Audio-Language Models
• TAD: Token-Adaptive Contrastive Decoding with Confidence-Guided Gating for Hallucination Mitigation in Large Audio-Language Models
• Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation
• Contrastive Decoding: Open-ended Text Generation as Optimization
• RISE: Randomized Input Sampling for Explanation of Black-box Models
• Understanding Deep Networks via Extremal Perturbations and Smooth Masks

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems