Audio visual models struggle to explain answers with evidence
ANCHOR: Audio-Visually Grounded Chain-of-Thought Reasoning Benchmark
Computer Vision and Pattern Recognition
Summary
Sometimes computers can answer questions about videos and sounds but don’t explain their answers with the right clues from what they saw or heard. The authors created a new test called ANCHOR that checks if the computer’s explanation actually points to the right parts of a video, sounds, or places in the picture. They found that while computers often get the right answer, their explanations aren’t always well connected to the real evidence. This test helps show where and how these models need to improve in understanding and explaining what they see and hear.
What this means in practice
- •For multimedia ai developers: Evaluate and improve models for linking video and audio evidence to their answers for more trustworthy AI reasoning.
- •For video content analysis teams: Use precise test data to measure how well tools explain insights from video and audio streams by pointing to relevant evidence.
Authors
Joel Julin, Souraja Kundu, Liza Dahiya, George Z. Wei, Liuyue Xie, Jingru Yi, Xu Zhang, Hao Yang, Laszlo A. Jeni
Abstract
Multimodal large language models (MLLMs) perform well on audio-visual question answering, but correct answers do not necessarily imply that their stated reasoning is grounded in relevant perceptual evidence. Existing benchmarks mainly evaluate final-answer accuracy, offering limited insight into whether model-generated rationales correctly identify visual regions, temporal intervals, and audio events supporting predictions. We introduce ANCHOR, a benchmark and evaluation framework for measuring grounded-rationale alignment in multimodal reasoning, comprising 2,319 videos and 8,272 question-answer pairs with structured reasoning traces, temporal evidence, spatial bounding boxes, and salient audio cues, including a human-refined test set of 224 videos and 975 question-answer pairs. We separately measure answer correctness, hallucination, logical divergence, and conciseness, with direct spatial and temporal grounding measurements on a focused subset. Across the full benchmark, twelve contemporary MLLMs show a substantial alignment gap: models with comparable answer performance can differ markedly in whether their explanations are supported by annotated evidence. On a matched subset of 100 videos and 446 question-answer pairs, rankings remain highly stable across human-refined, Gemini-generated, and GPT-generated references (mean Pearson r=0.99), providing evidence against model-reference circularity within this subset. Grounding-cue ablations of three MLLMs on the full human-refined test set reveal a consistent modality hierarchy: removing video reduces accuracy by 38.4-41.8 points, while removing bounding boxes (3.2-7.7 points) and audio (4.2-5.6 points) causes smaller, complementary, model-dependent drops. These findings show answer-only evaluation obscures important grounding failures and position ANCHOR as a testbed for evaluating accurate and evidence-aligned audio-visual reasoning.