Benchmark Test-Time Scaling of General LLM Agents

TL;DR

This paper introduces General AgentBench, a comprehensive benchmark for evaluating general-purpose LLM agents across multiple domains, revealing significant performance challenges and limitations of current scaling methods.

Contribution

The paper presents a new unified benchmark for testing general LLM agents and systematically analyzes their scaling behaviors and limitations in realistic, multi-skill environments.

Findings

Performance drops when moving from domain-specific to general settings

Neither sequential nor parallel scaling improves performance effectively

Fundamental limitations include context ceiling and verification gap

Abstract

LLM agents are increasingly expected to function as general-purpose systems capable of resolving open-ended user requests. While existing benchmarks focus on domain-aware environments for developing specialized agents, evaluating general-purpose agents requires more realistic settings that challenge them to operate across multiple skills and tools within a unified environment. We introduce General AgentBench, a benchmark that provides such a unified framework for evaluating general LLM agents across search, coding, reasoning, and tool-use domains. Using General AgentBench, we systematically study test-time scaling behaviors under sequential scaling (iterative interaction) and parallel scaling (sampling multiple trajectories). Evaluation of ten leading LLM agents reveals a substantial performance degradation when moving from domain-specific evaluations to this general-agent setting. Moreover, we find that neither scaling methodology yields effective performance improvements in practice, due to two fundamental limitations: context ceiling in sequential scaling and verification gap in parallel scaling. Code is publicly available at https://github.com/cxcscmu/General-AgentBench.