Multi-Agent System

Linking Process to Outcome: Conditonal Reward Modeling for LLM Reasoning

Aegis: Automated Error Generation and Identification for Multi-Agent Systems

Simulating Human-like Daily Activities with Desire-driven Autonomy

Amulet: ReAlignment During Test Time for Personalized Preference Adaptation of LLMs

In-Context Editing: Learning Knowledge from Self-Induced Distributions