Yunjue Agent Tech Report: A Fully Reproducible, Zero-Start In-Situ Self-Evolving Agent System for Open-Ended Tasks

2026-01-26

AI Summary

Traditional agents in open environments are constrained by static toolsets and offline training, struggling to adapt to task distribution shifts and scarce supervision. This work proposes an in-situ self-evolution paradigm that treats task interaction as a continuous stream of experience, enabling unsupervised distillation and reuse of capabilities through execution feedback, with tool evolution as the core mechanism for continual capability expansion from scratch. We design a parallelized batch evolution strategy to enhance efficiency and introduce a novel metric to monitor evolutionary convergence. Evaluated across five diverse benchmarks, our system achieves performance from scratch that significantly surpasses closed-source baselines; warm-start experiments further demonstrate seamless transfer of its generalizable knowledge to new domains. Code, system trajectories, and evolved tools are publicly released.

Technology Category

• Machine Learning: Evolutionary Learning
• Multiagent Systems: Adversarial Agents
• Search and Optimization: Evolutionary Computation

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Web Mining and Content Analysis: Models for Web evolution
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Conventional agent systems often struggle in open-ended environments where task distributions continuously drift and external supervision is scarce. Their reliance on static toolsets or offline training lags behind these dynamics, leaving the system's capability boundaries rigid and unknown. To address this, we propose the In-Situ Self-Evolving paradigm. This approach treats sequential task interactions as a continuous stream of experience, enabling the system to distill short-term execution feedback into long-term, reusable capabilities without access to ground-truth labels. Within this framework, we identify tool evolution as the critical pathway for capability expansion, which provides verifiable, binary feedback signals. Within this framework, we develop Yunjue Agent, a system that iteratively synthesizes, optimizes, and reuses tools to navigate emerging challenges. To optimize evolutionary efficiency, we further introduce a Parallel Batch Evolution strategy. Empirical evaluations across five diverse benchmarks under a zero-start setting demonstrate significant performance gains over proprietary baselines. Additionally, complementary warm-start evaluations confirm that the accumulated general knowledge can be seamlessly transferred to novel domains. Finally, we propose a novel metric to monitor evolution convergence, serving as a function analogous to training loss in conventional optimization. We open-source our codebase, system traces, and evolved tools to facilitate future research in resilient, self-evolving intelligence.

Problem

• open-ended tasks
• task distribution drift
• external supervision scarcity
• static toolsets
• capability boundaries

Innovation

• In-Situ Self-Evolving
• Tool Evolution
• Zero-Start Agent
• Parallel Batch Evolution
• Open-Ended Tasks