AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?

2026-07-31

AI Summary

Current evaluations of self-evolving large language model agents are largely confined to isolated tasks, failing to capture their behavioral characteristics in realistic streaming scenarios. This work proposes AgentStream, the first evaluation paradigm tailored for streaming tasks, which constructs configurable task streams to systematically assess the performance of five representative self-evolution methods combined with three state-of-the-art foundation models under three distinct settings: isolated, sequential, and interleaved. The study reveals that the efficacy of self-evolution is jointly influenced by model capability, method architecture, and task stream dynamics: its reliability varies across scenarios, performance gains exhibit diminishing returns and are non-monotonic, and no single method consistently outperforms others across all configurations.

Abstract

Large language model (LLM) agents can self-evolve by continually improving from their own accumulated experience. However, existing studies predominantly adopt independent evaluation. Consequently, the behavior of self-evolving agents in realistic streaming settings, where agents adapt to diverse and complex task streams, remains poorly understood. To address this gap, we introduce AgentStream, a unified framework that evaluates self-evolving agents spanning diverse evolution components by organizing agentic benchmarks into a configurable task stream and instantiating the \texttt{Isolated}, \texttt{Sequential}, and \texttt{Interleaved} streaming scenarios at test time, which progressively vary the scope and domain composition of the stream. Over these scenarios, we combinatorially evaluate five representative self-evolving methods across three frontier foundation models, disentangling how model capability, method architecture, and streaming scenario jointly shape self-evolution. Our results show that self-evolution reliability varies across streaming scenarios, the benefit of self-evolution is gated by model capability and non-monotonic in model strength, and no single method dominates across models and scenarios. These findings offer concrete guidance for selecting self-evolving methods across models and streaming scenarios. Overall, we advocate that self-evolving agents should be evaluated under realistic task streams rather than isolated single-task settings.

Problem

• self-evolving agents
• streaming tasks
• LLM agents
• task streams
• evaluation gap

Innovation

• self-evolving agents
• streaming tasks
• AgentStream
• LLM evaluation
• task streams