analiza aiqbench
SocialRL introduces a multi-turn reinforcement learning framework that successfully propagates delayed rewards to improve long-horizon planning in autonomous agents. By dynamically balancing goal advancement with relational attunement across dialogue stages, the method achieves significant gains in social-dialogue benchmarks without relying on external hardware or quantum principles. This work represents a substantive step forward in refining the social intelligence capabilities of large language models for human-AI collaboration.
SocialRL: Refining LLMs' Social Intelligence through Multi-turn Reinforcement Learning and Reward Design
arXiv:2609.09764v1 Announce Type: new Abstract: Social intelligence enables agents to read social context, infer intent, and adapt over sustained dialogue. As language models become autonomous collaborators, it is central to building effective and trustworthy human-AI interaction. Existing reinforcement learning methods optimize single-turn utterances and sparse outcome rewards, producing short-sighted policies that struggle to manage goal-relationship tensions across multi-turn interactions. We propose SocialRL, a multi-turn reinforcement learning framework addressing both challenges. First, we apply multi-turn reinforcement learning using PPO that propagates delayed outcome rewards back to each turn, enabling long-horizon planning. Second, we design six process reward dimensions capturing the goal-relationship trade-off, including goal advancement, relational attunement, contextual coherence, etc. A reward model dynamically generates fine-grained scoring criteria for each dimension, while a stage-aware weight schedule prioritizes relationship-building in early turns, goal advancement mid-way, and balanced closure late. Across multiple social-dialogue benchmarks, SocialRL improves Goal Achievement by an average of 9.2 percentage points over the corresponding Base models. These results demonstrate the effectiveness of SocialRL across synthetic and real social scenes, as well as standard and challenging social scenarios.
Abstract:Social intelligence enables agents to read social context, infer intent, and adapt over sustained dialogue. As language models become autonomous collaborators, it is central to building effective and trustworthy human-AI interaction. Existing reinforcement learning methods optimize single-turn utterances and sparse outcome rewards, producing short-sighted policies that struggle to manage goal-relationship tensions across multi-turn interactions. We propose SocialRL, a multi-turn reinforcement learning framework addressing both challenges. First, we apply multi-turn reinforcement learning using PPO that propagates delayed outcome rewards back to each turn, enabling long-horizon planning. Second, we design six process reward dimensions capturing the goal-relationship trade-off, including goal advancement, relational attunement, contextual coherence, etc. A reward model dynamically generates fine-grained scoring criteria for each dimension, while a stage-aware weight schedule prioritizes relationship-building in early turns, goal advancement mid-way, and balanced closure late. Across multiple social-dialogue benchmarks, SocialRL improves Goal Achievement by an average of 9.2 percentage points over the corresponding Base models. These results demonstrate the effectiveness of SocialRL across synthetic and real social scenes, as well as standard and challenging social scenarios.
Computation and Language (cs.CL)
arXiv:2609.09764 [cs.CL]
(or arXiv:2609.09764v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.09764 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 268 cuvinte
Subiecte conexe
State of the Field: AI for Epistemics and Coordination
LessWrong · 2 oct. 2026
PsychJail: Exploring Psychological Jailbreaks via Multi-Turn Persuasion of LLM Policies
arXiv · 25 aug. 2026
Learning to GUIDE the Group: Self-Learning LLM Facilitation Across Multi-Party Tasks
arXiv · 15 oct. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026