large-audio-language-models-multiple-choice-audio-reasoning-benchmarks-mmar CLM-0410observationsingle paperReviewed by AI
For large audio language models on multiple-choice audio reasoning benchmarks (MMAR, MMAU-mini), accuracy does not track whether the model's chain of thought is acoustically grounded or causally drives the answer: several models answer correctly from perception steps that assert unsupported labels, and for some models replacing the reasoning chain with vacuous or fabricated content leaves the final answer unchanged.
Capability: Understanding speech and audio
Observed on
Nine LALMs (open and closed source) on two 1,000-item multiple-choice audio reasoning benchmarks, with reasoning externalized via a fixed four-step CoT prompt..
Sources
- Validated by: The six-metric mean correlated with 3-rater human CoT quality ratings at Spearman 0.671 (p=0.006) over 450 sampled instances, above every single metric (best M1, 0.563) and every two-metric subgroup; rankings also matched two LLM judges and CoT-perturbation-derived mode assignments, and were stable across datasets (model identity explained 91.9-96.4% of variance) and across scorer backbones.. Measured: per-metric scores and accuracy for 9 models on 2 benchmarks, plus a manual CoT-perturbation check on randomly selected instances and 450 human-rated CoTs. [truncated]
Either way, bring evidence: a link, or what you saw.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- On long-video (10-150 minute) audio-visual reasoning questions, multiple-choice accuracy overstates grounding: a large open model can answer about 55% of items correctly while scoring 8.5 on spatio-temporal evidence localization, whereas a frontier proprietary model at 73% accuracy scores 35.2 on the same evidence metric, so answer accuracy alone cannot distinguish deduction from prior-driven guessing.Reasoning about time in video · reviewed by AI
- For spoken language models answering questions over long dialogue histories, converting the history to transcripts beats keeping it as audio when the answer is a lexical detail (exact slot values across dense turns: mean 71.3% vs 42.6%), but loses when the answer depends on how something was said, where audio history wins for six of seven models and by a larger margin once lexical emotion cues are removed.Understanding speech and audio · reviewed by AI
- On multi-session dialogue questions whose answers depend only on how an utterance was delivered, a cascaded ASR-then-LLM pipeline (Whisper large-v3 into a frontier model) scores below 7B audio-native models on the same clips, and below even a transcript-only baseline, because the transcription step drops delivery cues entirely and adds content errors.Understanding speech and audio · reviewed by AI
- In audio language models that feed encoder outputs into a text LLM, downstream understanding accuracy across speech, sound and music is set by how semantically text-aligned the encoder is rather than by whether its representation is discrete or continuous — reconstruction-oriented codecs with high acoustic fidelity underperform, semantically distilled tokenizers can beat their own continuous teacher, and scaling the LLM backbone does not recover the lost information.Understanding speech and audio · reviewed by AI
- In closed-book multi-hop QA the accuracy gain from decomposed prompting shrinks to little in frontier models, but answers still change between task-equivalent prompt regimes (agreement only about 70%, and about 60% on MuSiQue), and because stable facts survive re-prompting while hallucinations do not, disagreement across regimes detects errors better than standard uncertainty baselines on F1 and AUROC without retrieval or training.Recalling facts it was trained on · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: Show that on these benchmarks accuracy rankings match process-quality rankings, and that perturbing the CoT changes the final answer for all tested models, i.e. no model answers correctly with a non-causal or ungrounded chain.