The paper presents a new evaluation protocol for post‑training quantization of speech language models that separates lexical output, transcript‑insufficient endpoints, and packed implementations. In a Qwen2‑Audio case study, a 6‑bit allocation selected for translation improves chrF scores but degrades emotion recognition, while uniform and front‑layer controls perform better on emotion tasks. Similar patterns hold at 7 bits, and a 4‑bit study shows consistent emotion deficits across all low‑bit allocations, with no advantage for the selected scheme. The study highlights a precision‑dependent mismatch between lexical output, waveform‑dependent behavior, and nominal precision, without claiming a general failure of low‑bit models or a deployment benefit for the selected allocation.
By Mengzhe Geng, Jinxi Jin, Junhao Xu
The paper introduces the Speech-Unsupported Rejection Evaluation Challenge (SURE‑Challenge), a benchmark designed to test whether speech‑LLMs should accept or reject audio inputs before generating answers. Using LibriSpeech‑derived transcriptions paired with first‑word question answering, the authors evaluate various noise and silence conditions, and compare a simple energy‑plus‑Whisper‑score rule against a Qwen2‑Audio front‑end. On a 474‑row test set, the rule rejects 196 of 204 unsupported inputs while preserving accuracy on supported data, revealing a pre‑generation error mode that answer‑only scoring misses.
By Mengzhe Geng
arXiv:2608.27783v3 Announce Type: replace-cross
Abstract: Speech language models (speech LLMs) can generate plausible outputs from audio that contains no usable speech evidence. We study this failure...
By Mengzhe Geng
The study evaluates Qwen2‑Audio‑7B‑Instruct quantization at mixed 4/8‑bit settings on English‑to‑German translation and emotion recognition tasks. BLEU and chrF scores relative to FP16 show no significant difference, yet absolute accuracy on RAVDESS drops by 3.71% for 6‑bit and 1.17% for 7‑bit, with the 6‑bit interval statistically significant. The results demonstrate that text‑output metrics alone cannot confirm performance preservation on speech tasks where labels cannot be inferred from transcripts.
By Mengzhe Geng, Jinxi Ji, Junhao Xu
arXiv:2609.18533v1 Announce Type: new
Abstract: Automatic speech recognition (ASR) systems exhibit unequal error rates across speaker groups, motivating interventions on their internal representation...
By Nicolas Bourrel, Abderrahmane Issam, Gerasimos Spanakis
The paper introduces a reproducible framework that converts existing text-based tool‑calling benchmarks into audio‑based evaluations using text‑to‑speech, speaker variation, and environmental noise, while preserving original annotations. Experiments on seven omni‑modal models show that performance varies strongly by model and task, with notable gaps between text and voice results. The study also includes a stress test, a reference‑free LLM‑as‑judge protocol, and demonstrates that open‑source judges can match proprietary ones, aiding privacy‑preserving evaluation.
By Md Tahmid Rahman Laskar, Xue-Yong Fu, Seyyed Saeed Sarfjoo, Quinten McNamara, Jonas Robertson, Shashi Bhushan TN
The paper investigates how to evaluate audio‑language models by separating the use of acoustic evidence from the need to invoke a generative audio model. Using a controlled call‑decision framework, the authors compare policies that rely on transcript labels, encoder outputs from CLAP, AST, or WavLM, and optional calls to generative models such as Qwen2‑Audio, Qwen2.5‑Omni, or MOSS‑Audio. Results on the VocalSound dataset show that while transcript‑only accuracy is low (0.296), encoder‑only controls achieve high accuracy (≈0.85) without any generative calls, and adding generative calls yields only a marginal improvement (0.925 vs. 0.921).
By Mengzhe Geng
arXiv:2609.20995v1 Announce Type: cross
Abstract: Natural spoken interaction requires more than streaming ASR, language generation, and speech synthesis: a system must react to overlap without cancel...
By Bertil Braun
VoxReason introduces a listener‑free evaluation framework that measures whether a speech planning system’s delivery choices—such as pitch, energy, rate, pause, emphasis, and stance—are grounded in cited source records before any waveform is generated. The system outputs a source‑cited speaking plan and uses a deterministic verifier to check citation legality, slot agreement, unsupported states, schema validity, and counterfactual locality. Experiments on 1,440 source‑label cases show that simple slot accuracy can be misleading, while a 7B locality‑based repair model significantly improves plan‑slot accuracy and locality, and removing source records sharply reduces the grounded score.
whyItMatters":"The framework provides a concrete, measurable way to ensure that expressive speech systems make source‑licensed planning decisions, addressing a source‑use failure that occurs before audio synthesis."
By Mengzhe Geng
The paper investigates how to evaluate generative audio large language models (Audio‑LLMs) on known closed‑set tasks by separating the decision to call a generative model from the use of acoustic evidence. It introduces a controlled call‑decision framework where a policy can choose between a transcript label, encoder evidence from CLAP, AST, or WavLM, or a generative call to Qwen2‑Audio, Qwen2.5‑Omni, or MOSS‑Audio, and measures the impact of generative calls on accuracy. Results on the VocalSound dataset show that while transcript‑only accuracy is low (0.296), encoder‑based controls achieve high accuracy (≈0.85) without any generative calls, and adding generative calls yields only a marginal improvement (0.925 vs. 0.921).
By Mengzhe Geng
Automatic speech recognition (ASR) systems exhibit unequal error rates across speaker groups, motivating interventions on their internal representations. We ask whether speaker-linked attributes that...
The study evaluates conversational speech models by placing them at 427 pauses and 482 turn ends from 176 unscripted conversations, scoring their responses on timing, latency, length, wording, and recall of recent dialogue. Hosted APIs (GPT Realtime, Gemini 3.1 Flash Live, Grok Voice) respond at almost every pause, with latency ranging from 1–4 s for silence-based endpointers and about 8 s for semantic turn detection, while full‑duplex models (PersonaPlex, Moshi) remain silent more often and can interrupt within human reaction time. Recall rates vary: hosted systems remember 41–60 % of facts under a Claude Opus 5.5 judge and 19–26 % with a strict word matcher; full‑duplex models recall 17–22 % and 12–15 % respectively.
By Arya Bastani, Iman Modarressi