# People Making Conversation: Evaluating How Naturally Speech Models Converse

arXiv: https://arxiv.org/abs/2610.12802 · HTML: https://mlsanity.com/paper/2610.12802/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 25 references that are in the map.
- First posted on arXiv 8 Oct 2026.

## Builds on (most cited first, 18 of 18)

- [Moshi: a speech-text foundation model for real-time dialogue](https://mlsanity.com/paper/2410.00037/) (2024)
- [VoiceBench: Benchmarking LLM-Based Voice Assistants](https://mlsanity.com/paper/2410.17196/) (2024)
- [Full-Duplex-Bench: A Benchmark to Evaluate Full-duplex Spoken Dialogue Models on Turn-taking Capabilities](https://mlsanity.com/paper/2503.04721/) (2025)
- [Beyond Turn-Based Interfaces: Synchronous LLMs as Full-Duplex Dialogue Agents](https://mlsanity.com/paper/2409.15594/) (2024)
- [PersonaPlex: Voice and Role Control for Full Duplex Conversational Speech Models](https://mlsanity.com/paper/2602.06053/) (2026)
- [Voice Activity Projection: Self-supervised Learning of Turn-taking Events](https://mlsanity.com/paper/2205.09812/) (2022)
- [Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics](https://mlsanity.com/paper/2503.01174/) (2025)
- [Full-Duplex-Bench-v2: A Multi-Turn Evaluation Framework for Duplex Dialogue Systems with an Automated Examiner](https://mlsanity.com/paper/2510.07838/) (2025)
- [Seamless Interaction: Dyadic Audiovisual Motion Modeling and Large-Scale Dataset](https://mlsanity.com/paper/2506.22554/) (2025)
- [Full-Duplex-Bench-v3: Benchmarking Tool Use for Full-Duplex Voice Agents Under Real-World Disfluency](https://mlsanity.com/paper/2604.04847/) (2026)
- [SALMONN-omni: A Standalone Speech LLM without Codec Injection for Full-duplex Conversation](https://mlsanity.com/paper/2505.17060/) (2025)
- [MTR-DuplexBench: Towards a Comprehensive Evaluation of Multi-Round Conversations for Full-Duplex Speech Language Models](https://mlsanity.com/paper/2511.10262/) (2025)
- [TurnBench: A Multi-Domain Benchmark for Turn-Taking Dynamics in Spoken Dialogue](https://mlsanity.com/paper/2608.25218/) (2026)
- [BayLing-Duplex: Native Full-Duplex Speech Dialogue with a Single Autoregressive LLM](https://mlsanity.com/paper/2606.14528/) (2026)
- [Instruct-FD: Can Your Full-Duplex Speech System Follow Turn-Taking Instructions?](https://mlsanity.com/paper/2607.20460/) (2026)
- [Speak or Stay Silent: Context-Aware Turn-Taking in Multi-Party Dialogue](https://mlsanity.com/paper/2603.11409/) (2026)
- [CASPER: A Large Scale Spontaneous Speech Dataset](https://mlsanity.com/paper/2506.00267/) (2025)
- [CoDeTT: A Context-Aware Decision Benchmark for Turn-Taking Evaluation](https://mlsanity.com/paper/2603.25434/) (2026)

## Abstract (arXiv)

A conversational speech model has to decide, at every pause, whether to speak. Existing evaluations test boundary detection, policy adherence, or aggregate statistics, not this decision. We place real-time speech systems at 427 pauses and 482 turn ends from 176 unscripted conversations in Seamless Interaction, give each the conversation so far, and score what it generates against what the person did: whether and how soon it speaks at a pause; its latency, length, and wording at a turn end; and whether it can answer a question about the preceding 30 to 120 seconds. Across six systems, the hosted APIs (GPT Realtime, Gemini 3.1 Flash Live, Grok Voice) respond at almost every pause whichever endpointer they use; only their latency, about 1 to 4 s with silence-based endpointers and about 8 s with semantic turn detection, decides whether the words land inside the pause. The full-duplex models (PersonaPlex, Moshi) are the only systems that regularly stay silent and the only ones that interrupt within a human reaction time. Hosted systems reply longer than the person. Given the history as text, hosted systems recall 41 to 60% of conversation facts under a Claude Opus 5.5 judge and 19 to 26% under a strict word matcher; hearing it as audio, full-duplex models recall 17 to 22% and 12 to 15%. We release the labels, points, harness, and every output.