AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications
This work is significant for developers and researchers working on LLM-based autonomous agents, as it provides a more realistic and scalable evaluation framework for long-horizon memory, which is critical for complex agentic applications.
This paper introduces AMA-Bench, a new benchmark for evaluating long-horizon memory in LLM-based autonomous agents, addressing the gap between existing dialogue-centric benchmarks and real-world agent-environment interactions. The benchmark includes real-world and synthetic agentic trajectories with expert-curated and rule-based QA. The authors also propose AMA-Agent, a memory system that achieves 57.22% average accuracy on AMA-Bench, outperforming existing baselines by 11.16%.
Large Language Models (LLMs) are deployed as autonomous agents in increasingly complex applications, where enabling long-horizon memory is critical for achieving strong performance. However, a significant gap exists between practical applications and current evaluation standards for agent memory: existing benchmarks primarily focus on dialogue-centric, human-agent interactions. In reality, agent memory consists of a continuous stream of agent-environment interactions that are primarily composed of machine-generated representations. To bridge this gap, we introduce AMA-Bench (Agent Memory with Any length), which evaluates long-horizon memory for LLMs in real agentic applications. It features two key components: (1) a set of real-world agentic trajectories across representative agentic applications, paired with expert-curated QA, and (2) a set of synthetic agentic trajectories that scale to arbitrary horizons, paired with rule-based QA. Our comprehensive study shows that existing memory systems underperform on AMA-Bench primarily because they lack causality and objective information and are constrained by the lossy nature of similarity-based retrieval employed by many memory systems. To address these limitations, we propose AMA-Agent, an effective memory system featuring a causality graph and tool-augmented retrieval. Our results demonstrate that AMA-Agent achieves 57.22% average accuracy on AMA-Bench, surpassing the strongest memory system baselines by 11.16%.