Agentic Knowledgeable Self-awareness
Shuofei Qiao, Zhisong Qiu, Baochang Ren, Xiaobin Wang, Xiangyuan Ru, Ningyu Zhang, Xiang Chen, Yong Jiang, Pengjun Xie, Fei Huang, Huajun Chen
2025-04-07
Summary
This paper talks about teaching AI to be self-aware like humans during decision-making, so it knows when to use its knowledge instead of always relying on everything it’s learned.
What's the problem?
Current AI helpers use all their knowledge for every task, like flooding a garden with water everywhere, which wastes resources and can lead to mistakes.
What's the solution?
They made a system called KnowSelf that trains AI to add special markers to its decisions, helping it choose the right amount of knowledge for each situation, like watering only where plants need it.
Why it matters?
This makes AI assistants smarter and more efficient, saving energy and improving decisions in apps like customer service or tutoring by focusing on what’s needed.
Abstract
Large Language Models (LLMs) have achieved considerable performance across various agentic planning tasks. However, traditional agent planning approaches adopt a "flood irrigation" methodology that indiscriminately injects gold trajectories, external feedback, and domain knowledge into agent models. This practice overlooks the fundamental human cognitive principle of situational self-awareness during decision-making-the ability to dynamically assess situational demands and strategically employ resources during decision-making. We propose agentic knowledgeable self-awareness to address this gap, a novel paradigm enabling LLM-based agents to autonomously regulate knowledge utilization. Specifically, we propose KnowSelf, a data-centric approach that applies agents with knowledgeable self-awareness like humans. Concretely, we devise a heuristic situation judgement criterion to mark special tokens on the agent's self-explored trajectories for collecting training data. Through a two-stage training process, the agent model can switch between different situations by generating specific special tokens, achieving optimal planning effects with minimal costs. Our experiments demonstrate that KnowSelf can outperform various strong baselines on different tasks and models with minimal use of external knowledge. Code is available at https://github.com/zjunlp/KnowSelf.