AgentLongBench: A Controllable Long Benchmark For Long-Contexts Agents via Environment Rollouts

TL;DR

AgentLongBench introduces a dynamic, environment-based benchmark for evaluating long-context agents, revealing their struggles with information synthesis in complex, knowledge-intensive scenarios.

Contribution

It presents a novel environment rollout framework for long-context evaluation, addressing static limitations of existing benchmarks and exposing new challenges for large language models.

Findings

State-of-the-art models excel at static retrieval but struggle with dynamic information synthesis.

Performance degradation correlates with the minimum token count needed to resolve queries.

High information density in tool responses significantly challenges agent capabilities.

Abstract

The evolution of Large Language Models (LLMs) into autonomous agents necessitates the management of extensive, dynamic contexts. Current benchmarks, however, remain largely static, relying on passive retrieval tasks that fail to simulate the complexities of agent-environment interaction, such as non-linear reasoning and iterative feedback. To address this, we introduce \textbf{AgentLongBench}, which evaluates agents through simulated environment rollouts based on Lateral Thinking Puzzles. This framework generates rigorous interaction trajectories across knowledge-intensive and knowledge-free scenarios. Experiments with state-of-the-art models and memory systems (32K to 4M tokens) expose a critical weakness: while adept at static retrieval, agents struggle with the dynamic information synthesis essential for workflows. Our analysis indicates that this degradation is driven by the minimum number of tokens required to resolve a query. This factor explains why the high information density inherent in massive tool responses poses a significantly greater challenge than the memory fragmentation typical of long-turn dialogues.