Edaizi/EvolveR-NQ-HotpotQA
This repository contains the dataset associated with the paper EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle. EvolveR is a framework designed to enable LLM agents to self-improve through a complete, closed-loop experience lifecycle. This lifecycle comprises two key stages: (1) Offline Self-Distillation, where interaction trajectories are synthesized into reusable principles, and (2) Online Interaction, where the agent retrieves these principles to guide… See the full description on the dataset page: https://huggingface.co/datasets/Edaizi/EvolveR-NQ-HotpotQA.
This repository contains the dataset associated with the paper EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle.
EvolveR is a framework designed to enable LLM agents to self-improve through a complete, closed-loop experience lifecycle. This lifecycle comprises two key stages: (1) Offline Self-Distillation, where interaction trajectories are synthesized into reusable principles, and (2) Online Interaction, where the agent retrieves these principles to guide decision-making. The effectiveness of this approach is demonstrated on complex multi-hop question-answering benchmarks.