# DiscussLLM: Teaching Large Language Models When to Speak

arXiv: https://arxiv.org/abs/2508.18167 · HTML: https://mlsanity.com/paper/2508.18167/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 42 references that are in the map.
- First posted on arXiv 25 Aug 2025.

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

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Scaling Synthetic Data Creation with 1,000,000,000 Personas](https://mlsanity.com/paper/2406.20094/) (2024)
- [VideoLLM-online: Online Video Large Language Model for Streaming Video](https://mlsanity.com/paper/2406.11816/) (2024)
- [Proactive Agent: Shifting LLM Agents from Reactive Responses to Active Assistance](https://mlsanity.com/paper/2410.12361/) (2024)
- [Streaming Dense Video Captioning](https://mlsanity.com/paper/2404.01297/) (2024)
- [A Survey on Proactive Dialogue Systems: Problems, Methods, and Prospects](https://mlsanity.com/paper/2305.02750/) (2023)
- [Synthetic Dialogue Dataset Generation using LLM Agents](https://mlsanity.com/paper/2401.17461/) (2024)
- [Proactive Conversational Agents with Inner Thoughts](https://mlsanity.com/paper/2501.00383/) (2024)
- [Synergi: A Mixed-Initiative System for Scholarly Synthesis and Sensemaking](https://mlsanity.com/paper/2308.07517/) (2023)
- [SocialMind: LLM-based Proactive AR Social Assistive System with Human-like Perception for In-situ Live Interactions](https://mlsanity.com/paper/2412.04036/) (2024)
- [A Survey on Recent Advances in Conversational Data Generation](https://mlsanity.com/paper/2405.13003/) (2024)
- [CodingGenie: A Proactive LLM-Powered Programming Assistant](https://mlsanity.com/paper/2503.14724/) (2025)
- [OmniChat: Enhancing Spoken Dialogue Systems with Scalable Synthetic Data for Diverse Scenarios](https://mlsanity.com/paper/2501.01384/) (2025)
- [Diverse and Effective Synthetic Data Generation for Adaptable Zero-Shot Dialogue State Tracking](https://mlsanity.com/paper/2405.12468/) (2024)
- [Redefining Proactivity for Information Seeking Dialogue](https://mlsanity.com/paper/2410.15297/) (2024)
- [A Framework for Synthetic Audio Conversations Generation using Large Language Models](https://mlsanity.com/paper/2409.00946/) (2024)
- [Streaming Detection of Queried Event Start](https://mlsanity.com/paper/2412.03567/) (2024)
- [ConvoGen: Enhancing Conversational AI with Synthetic Data: A Multi-Agent Approach](https://mlsanity.com/paper/2503.17460/) (2025)
- [Artificial Conversations, Real Results: Fostering Language Detection with Synthetic Data](https://mlsanity.com/paper/2503.24062/) (2025)

## Built on by (most cited first, 2 of 2)

- [When2Speak: A Dataset for Temporal Participation and Turn-Taking in Multi-Party Conversations for Large Language Models](https://mlsanity.com/paper/2605.05626/) (2026)
- [Jarvis: A Proactive Speech Agent for Multi-Party Conversations](https://mlsanity.com/paper/2610.07506/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) have demonstrated remarkable capabilities in understanding and generating human-like text, yet they largely operate as reactive agents, responding only when directly prompted. This passivity creates an "awareness gap," limiting their potential as truly collaborative partners in dynamic human discussions. We introduce textitDiscussLLM, a framework designed to bridge this gap by training models to proactively decide not just textitwhat to say, but critically, textitwhen to speak. Our primary contribution is a scalable two-stage data generation pipeline that synthesizes a large-scale dataset of realistic multi-turn human discussions. Each discussion is annotated with one of five intervention types (e.g., Factual Correction, Concept Definition) and contains an explicit conversational trigger where an AI intervention adds value. By training models to predict a special silent token when no intervention is needed, they learn to remain quiet until a helpful contribution can be made. We explore two architectural baselines: an integrated end-to-end model and a decoupled classifier-generator system optimized for low-latency inference. We evaluate these models on their ability to accurately time interventions and generate helpful responses, paving the way for more situationally aware and proactive conversational AI.