Post-Training Recipe, More Than Model Family, Shapes Multi-Agent LLM Conversational Behavior
On the map: LLM agents and tool use → Social simulation and multi-agent debate
Abstract (arXiv)
Multi-LLM systems use multiple language models to deliberate, judge each other's outputs, or coordinate as agents. Their value depends on the models producing measurably different conversational behaviors when given the same input. Prior offline studies recommend drawing one model per family for behavioral diversity, because LLMs prefer outputs from their own family when rating one another in isolation. Whether the same family label predicts behavior in interactive multi-LLM systems, the setting that real deployed systems use, has not been tested. We study this with a 940,000-chain 11-checkpoint corpus and a 1.6M-chain same-base Llama factorial. On our validated headline metric, hedging, a reasoning-distilled Llama checkpoint shifts by 18% depending on which same-base partner it replies to, more than any cross-family hedging gap in the controlled subset. Qwen, closed-API, and runtime checks suggest the pattern is not isolated, while repair and challenge analyses remain exploratory because their surface-cue detectors are weaker. Overall, the results identify post-training recipe as a first-class axis for multi-LLM panel composition and show that model family alone is an incomplete proxy for conversational diversity.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use. Builds on 15 of its 37 references that are in the map. First posted on arXiv 30 May 2026.
Builds on
15 of 37 references.
Closest ancestors · 2
Same direction, newest first
- 2023Simulating Opinion Dynamics with Networks of LLM-based Agents16 Nov
- 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate30 May
Other references · 13
- 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
- 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,988
- 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,784
- 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 817
- 2024LLM Evaluators Recognize and Favor Their Own Generationscited by 686
- 2023State of What Art? A Call for Multi-Prompt LLM Evaluationcited by 310
- 2024Self-Preference Bias in LLM-as-a-Judgecited by 258
- 2023Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2cited by 229
- 2024Large Language Models are Inconsistent and Biased Evaluatorscited by 153
- 2024Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinementcited by 114
- 2025Great Models Think Alike and this Undermines AI Oversightcited by 58
- 2024AI-AI Bias: large language models favor communications generated by large language modelscited by 26
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026"Who Am I, and Who Else Is Here?" Behavioral Differentiation Without Role Assignment in Multi-Agent LLM Systems
- 2025Evaluating LLM-based Agents for Multi-Turn Conversations: A Survey
- 2025Are LLM Agents Behaviorally Coherent? Latent Profiles for Social Simulation
- 2026Found in Conversation: LLMs Teach Themselves to Close the Multi-Turn Gap
- 2026Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets
- 2026Disentangling Topology and Diversity in Multi-Agent LLMs for Multilingual Low-Resource Emotion Detection
- 2025Agent-Centric Projection of Prompting Techniques and Implications for Synthetic Training Data for Large Language Models
- 2025Beyond Single Models: Enhancing LLM Detection of Ambiguity in Requests through Debate