Aligning Dialogue Agents with Global Feedback via Large Language Model Multimodal Reward Decomposition

Dong Won Lee, Hae Won Park, Cynthia Breazeal, Louis-Philippe Morency


Abstract
We propose a large language model based reward decomposition framework for aligning dialogue agents using only a single session-level feedback signal. We leverage the reasoning capabilities of a frozen, pretrained large language model (LLM) to infer fine-grained local implicit rewards by decomposing global, session-level feedback. Our first text-only variant prompts the LLM to perform reward decomposition using only the dialogue transcript. The second multimodal variant incorporates additional behavioral cues, such as pitch, gaze, and facial affect, expressed as natural language descriptions. These inferred turn-level rewards are distilled into a lightweight reward model, which we utilize for RL-based fine-tuning for dialogue generation. We evaluate both text-only and multimodal variants against state-of-the-art reward decomposition methods and demonstrate notable improvements in human evaluations of conversation quality, suggesting that LLMs are strong reward decomposers that obviate the need for manual reward shaping and granular human feedback.
Anthology ID:
2025.findings-emnlp.1239
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2025
Month:
November
Year:
2025
Address:
Suzhou, China
Editors:
Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, Violet Peng
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
22755–22787
Language:
URL:
https://aclanthology.org/2025.findings-emnlp.1239/
DOI:
Bibkey:
Cite (ACL):
Dong Won Lee, Hae Won Park, Cynthia Breazeal, and Louis-Philippe Morency. 2025. Aligning Dialogue Agents with Global Feedback via Large Language Model Multimodal Reward Decomposition. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 22755–22787, Suzhou, China. Association for Computational Linguistics.
Cite (Informal):
Aligning Dialogue Agents with Global Feedback via Large Language Model Multimodal Reward Decomposition (Lee et al., Findings 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.findings-emnlp.1239.pdf
Checklist:
 2025.findings-emnlp.1239.checklist.pdf