ComplLLM: Fine-Tuning LLMs to Discover Complementary Signals for Decision-Making
Conference on Neural Information Processing Systems (NeurIPS) 2026
ComplLLM fine-tunes a decision-assistant LLM to emit signals that complement a recommending agent's output, which a supervisor agent combines with its own information to reach a decision.
Abstract
Multi-agent decision pipelines can outperform single agent workflows when complementarity holds, i.e., different agents bring unique information to the table to inform a final decision. We propose ComplLLM, a post-training framework based on decision theory that fine-tunes a decision-assistant LLM using complementary information as reward to output signals that complement existing agent decisions. We validate ComplLLM on synthetic and real-world tasks involving domain experts, demonstrating how the approach recovers known complementary information and produces plausible explanations of complementary signals to support downstream decision-makers.
Citation
BibTeX
@inproceedings{guo2026complllm,
title={ComplLLM: Fine-Tuning LLMs to Discover Complementary Signals for Decision-Making},
author={Guo, Ziyang and Wu, Yifan and Hartline, Jason and Holstein, Kenneth and Hullman, Jessica},
booktitle={Advances in Neural Information Processing Systems},
year={2026}
}