跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.22627 · v1

Beyond Mimicry: Constrained Exploration with GRPO for Joint Multi-Talker ASR and Diarization under Unknown Speaker Counts

首次提交 2026-11-17 19:45
版本更新 2026-11-17 19:45
雷达观测 2026-11-18 11:33
arXiv ↗PDF ↗

摘要

Joint multi-talker ASR with speaker attribution and timestamps is particularly challenging when the speaker count is unknown and the output must follow a strict structure. We observe that Supervised Fine-Tuning (SFT)-only alignment of SpeechLLMs is brittle under dense overlaps, often leading to burst hallucinations and malformed speaker/timestamp tags. We propose a unified two-stage generative framework: (i) CoT-augmented SFT that infers the global speaker count before transcription, and (ii) GRPO-based constrained exploration guided by a Multi-dimensional Constraint-Aware Reward (MCAR) that directly optimizes permutation-invariant accuracy while enforcing counting, temporal, and structural constraints. On Libri3Mix and a Dynamic-Mix (2+3) set (speaker count unknown at test time), our model achieves 14.52%/9.24% cpWER and 1.95%/1.12% WDER, respectively, with GRPO yielding a 35%/54% relative cpWER reduction over SFT+CoT.

作者备注:5 pages, Interspeech 2026

期刊信息:Proc. Interspeech 2026, pp. 4587-4591

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-17 19:45 · 最新观测版本

Beyond Mimicry: Constrained Exploration with GRPO for Joint Multi-Talker ASR and Diarization under Unknown Speaker Counts

Joint multi-talker ASR with speaker attribution and timestamps is particularly challenging when the speaker count is unknown and the output must follow a strict structure. We observe that Supervised Fine-Tuning (SFT)-only alignment of SpeechLLMs is brittle under dense overlaps, often leading to burst hallucinations and malformed speaker/timestamp tags. We propose a unified two-stage generative framework: (i) CoT-augmented SFT that infers the global speaker count before transcription, and (ii) GRPO-based constrained exploration guided by a Multi-dimensional Constraint-Aware Reward (MCAR) that directly optimizes permutation-invariant accuracy while enforcing counting, temporal, and structural constraints. On Libri3Mix and a Dynamic-Mix (2+3) set (speaker count unknown at test time), our model achieves 14.52%/9.24% cpWER and 1.95%/1.12% WDER, respectively, with GRPO yielding a 35%/54% relative cpWER reduction over SFT+CoT.

查看此版本解读 →查看此版本原文 ↗