# Talk2Agent: Benchmarking Voice Interfaces for Text Agents

arXiv: https://arxiv.org/abs/2609.38867 · HTML: https://mlsanity.com/paper/2609.38867/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 22 of its 31 references that are in the map.
- First posted on arXiv 30 Sep 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 22 of 22)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments](https://mlsanity.com/paper/2404.07972/) (2024)
- [SLURP: A Spoken Language Understanding Resource Package](https://mlsanity.com/paper/2011.13205/) (2020)
- [Deep context: end-to-end contextual speech recognition](https://mlsanity.com/paper/1808.02480/) (2018)
- [Semantic Distance: A New Metric for ASR Performance Analysis Towards Spoken Language Understanding](https://mlsanity.com/paper/2104.02138/) (2021)
- [HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models](https://mlsanity.com/paper/2309.15701/) (2023)
- [WildClawBench: A Benchmark for Real-World, Long-Horizon Agent Evaluation](https://mlsanity.com/paper/2605.10912/) (2026)
- [τ-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains](https://mlsanity.com/paper/2603.13686/) (2026)
- [STOP: A dataset for Spoken Task Oriented Semantic Parsing](https://mlsanity.com/paper/2207.10643/) (2022)
- [Can Contextual Biasing Remain Effective with Whisper and GPT-2?](https://mlsanity.com/paper/2306.01942/) (2023)
- [VoiceAgentBench: Are Voice Assistants ready for agentic tasks?](https://mlsanity.com/paper/2510.07978/) (2025)
- [EVA-Bench: A New End-to-end Framework for Evaluating Voice Agents](https://mlsanity.com/paper/2605.13841/) (2026)
- [Semantic-WER: A Unified Metric for the Evaluation of ASR Transcript for End Usability](https://mlsanity.com/paper/2106.02016/) (2021)
- [SeMaScore : a new evaluation metric for automatic speech recognition tasks](https://mlsanity.com/paper/2401.07506/) (2024)
- [Audio2Tool: Speak, Call, Act -- A Dataset for Benchmarking Speech Tool Use](https://mlsanity.com/paper/2604.22821/) (2026)
- [From Text to Voice: A Reproducible and Verifiable Framework for Evaluating Tool Calling LLM Agents](https://mlsanity.com/paper/2605.15104/) (2026)
- [DeRAGEC: Denoising Named Entity Candidates with Synthetic Rationale for ASR Error Correction](https://mlsanity.com/paper/2506.07510/) (2025)
- [Evaluation of Automatic Speech Recognition Using Generative Large Language Models](https://mlsanity.com/paper/2604.21928/) (2026)
- [AgenticASR: Refining Speech Recognition in Real-World Scenarios via an Agentic Approach](https://mlsanity.com/paper/2607.28175/) (2026)
- [How to Recognize New Words: A Comparison Between Context Biasing Methods and Speech LLMs](https://mlsanity.com/paper/2608.05759/) (2026)

## Abstract (arXiv)

Large language model (LLM) computer-use agents are typically evaluated with clean written instructions, despite speech being an increasingly popular interface for interacting with such systems. Speech input introduces an additional failure point: transcription errors can alter task-critical entities, constraints, or targets before the agent begins reasoning, while conventional ASR metrics do not directly measure whether the information required for successful execution has been preserved. We introduce Talk2Agent, a benchmark for evaluating how effectively voice interfaces convey human-spoken instructions to LLM-based computer-use agents. Talk2Agent builds human-spoken versions of tasks from WildClawBench and OSWorld and evaluates a range of voice interfaces, including dedicated ASR models, audio-capable LLMs, contextual biasing, and LLM-based ontology repair. Because repeatedly executing long-horizon computer-use tasks is costly and stochastic, we further propose an execution-free, task-conditioned evaluation framework that projects the original task grader onto prompt-addressable intentions and measures how much task-relevant information is retained after the voice interface. On WildClawBench, Talk2Agent's execution-free native projection provides a practical, execution-grounded measure of voice-interface quality, correlating with downstream task completion and improving Pearson correlation by 0.246 over WER/CER on 32 hours of real human speech.