arXiv Computation and Language

Auditing Generative Audio Calls for Known-Task Audio-LLM Evaluation

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).

arXiv Machine Learning
Sep 4

VoxReason: Listener-Free Evaluation of Source-Grounded Speech Planning Before Synthesis

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
arXiv Computation and Language
Aug 31

SURE-Challenge: Evaluating Speech Evidence Before Speech-LLM Generation

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
Hugging Face Trending Papers
Aug 20

Hear2Act: Benchmarking When Prosody Should Change What an Assistant Does

Prosodic cues can convey task-relevant information that alters the trajectory and outcome of a task-oriented dialogue, even when the words themselves remain unchanged. Yet existing benchmarks typically evaluate prosodic perception, response appropriateness, and task-oriented dialogue in isolation, making it difficult to test whether prosodic evidence changes downstream decisions.

arXiv Computation and Language
Sep 3

AVERT: Audio-Verified Adjudication for Spoken Dialogue State Tracking

AVERT is a method for spoken dialogue state tracking that improves upon a per-turn text editor by incorporating an audio-conditioned verifier to score candidate slot values. It addresses three types of recoverable errors—inconsistent values across turns, omitted slots, and values unsupported by audio—using three specialized operators: vote, add, and swap, each limited to relevant slots. On the SpokenWOZ dataset, AVERT achieves a joint goal accuracy of 40.13, surpassing both a base speech-LLM (33.04) and a text editor (38.34) without retraining, and matching the performance of a larger end‑to‑end system that processes the full spoken history.

By Chunggi Lee, Hanspeter Pfister
arXiv AI
Aug 28

SpeechGym: An Audio-Native Gym for Training Voice Agents via Reinforcement Learning

SpeechGym is an audio‑native environment that lets two omni‑modal models converse entirely in native audio, eliminating external ASR/TTS and API boundaries while preserving the tasks, tools, and success checks of a standard text‑based agent benchmark. By training end‑to‑end, the framework addresses perceptual failures—such as misheard arguments that cascade into failed calls—and behavioural failures, both of which are automatically labeled for free. Using per‑turn process rewards to overcome reward sparsity, agents trained in SpeechGym transfer to an independent voice benchmark, doubling task success and improving efficiency in turns and tokens.

By Jiajun Fan, Jingyuan Li, Prashanth Gurunath Shivakumar, Jia-Hong Huang, Qi Luo, M. Maruf, Ivan Bulyko, Ge Liu, Roger Ren
arXiv AI
6d ago

Scalable Context Orchestration for Serving LLMs Over Voice

The paper introduces llmovoice, a middleware that explicitly models voice context for large language model (LLM) serving in voice AI applications. By incorporating speaking rate, background noise, packet loss, and other paralinguistic factors into a bounded context, llmovoice guides the LLM to generate more aligned responses. Experiments show significant reductions in speaking‑rate errors, false interruptions, and model usage costs, especially in long voice sessions.

By Linyi Jiang, Silvery D. Fu, Yifei Zhu