Listen, Reason, and Segment: Aligning LALMs with Editorial Judgment for Media Chapterization
2026-08-17 • Sound
SoundArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors focus on a difficult but useful task called audio chapterization, which means dividing long audio recordings into meaningful sections, like chapters in a book. This is hard because chapter breaks depend on human judgment, not just obvious sounds. They created a new method named AudioChaps that improves chapterizing by training large audio language models with a special technique called Group Relative Policy Optimization, guided by step-by-step reasoning. They also made three new datasets to help train and test these models, and found that their method significantly outperforms previous models in accuracy. Their work shows promise for making long audio easier to navigate and organize.
Large Audio Language ModelsAudio ChapterizationGroup Relative Policy OptimizationChain-of-Thought ReasoningSupervised Fine-TuningAudio DatasetsF1 ScoreSequential ReasoningMedia IndexingContent Structuring
Authors
Tony Alex, Wish Suharitdamrong, Sara Atito, Armin Mustafa, Muhammad Awais, Philip J. B. Jackson, Jiankang Deng, Ismail Elezi
Abstract
Large Audio Language Models (LALMs) have made rapid progress on standardized benchmarks, yet their deployment in practical media workflows, curation, archival indexing, and content distribution remains largely unrealized. We identify automated audio chapterization, the task of segmenting continuous audio streams into thematically coherent chapters, as a demanding and commercially consequential setting that exposes this gap. Chapterization is challenging because boundaries are defined less by objective acoustic events than by subjective editorial judgment, requiring models to reason sequentially over long acoustic contexts and approximate creator-authored boundary decisions. We present AudioChaps, a post-training framework for aligning end-to-end LALMs for this task via Group Relative Policy Optimization (GRPO) guided by Chain-of-Thought (CoT) reasoning. To support training and evaluation, we curate three datasets: AudioChaps-Alignment, derived from creator-annotated chapter boundaries on YouTube; AudioChaps-CoT, which provides structured supervision for well-formatted, high-quality, and evidence-grounded boundary reasoning; and AudioChaps-Eval, a held-out benchmark for audio chapterization. Applying GRPO directly without a Supervised Fine-Tuning (SFT) cold start, AudioChaps-R1-Zero already improves average F1 by 33 points over the state-of-the-art LALM Audio-Flamingo-3-Think. The AudioChaps framework produces our final aligned LALM, AudioChaps-R1, which improves average F1 by 49 points. These results demonstrate that GRPO-trained LALMs can reliably transform unstructured auditory streams into navigable, structured media. Our code, models, and dataset resources will be released upon acceptance at https://github.com/ta012/AudioChaps.