This paper proposes FRESH, a Failure-aware Retrieval framework over Experience-Structured Heterogeneous graphs, which transforms historical successes and failures into structured external experience for tool-using agents and consistently improves task success and tool-use reliability over no-memory agents and representative memory-based baselines.
Abstract
Small and medium-sized language models offer cost-effective executors for tool-using agents, making them attractive for local and large-scale deployment. However, in long-horizon and stateful environments, they often make structural errors such as missing required observations, performing premature writes, repeating failed calls, and violating action preconditions. These errors can lead to incorrect state updates, policy violations, and costly or irreversible consequences, making reliable tool execution a critical deployment challenge. Existing fine-tuning approaches require substantial data and computation, while flat memory may retrieve failed actions without preserving their causal context or safety conditions. In this paper, we propose FRESH, a Failure-aware Retrieval framework over Experience-Structured Heterogeneous graphs, which transforms historical successes and failures into structured external experience for tool-using agents. By explicitly modeling the dependencies among tasks, actions, errors, repairs, and execution conditions, FRESH helps frozen language models reuse reliable strategies, avoid recurring failures, and make safer decisions in stateful tool interactions. Experiments on $\tau$-Bench and AppWorld with multiple open-source models show that FRESH consistently improves task success and tool-use reliability over no-memory agents and representative memory-based baselines.
The Adaptive Consistency Graph (ACG) is introduced, which incrementally organizes execution evidence and its provenance in a persistent graph, then constructs a temporary requirement-centered view for each decision under a bounded context budget.
This work studies insight-level memory maintenance for long-term language agents and proposes a failure-aware memory maintenance framework based on an editable insight graph and introduces a utility-aware retrieval mechanism and a graph controller that updates the memory graph after task execution.
NLPG is proposed, an external policy-memory method for improving a fixed agent without changing its model parameters or program structure that outperforms the strongest listed baseline for each benchmark by 8.71 percentage points on average.
Xu Liu, Wen-Zhang Wei, Jun Cao et al.· arXiv.org· 0 citations
Skill Runtime Intelligence is presented, a passive runtime-intelligence system that reconstructs supported Skill-lifecycle stages across heterogeneous harnesses while preserving unsupported stages as unknown and motivating executable adapter qualification.
A unified taxonomy of agent context compression along three dimensions is introduced: compression target (what is compressed), compression mechanism (how it is transformed and retained), and control policy (who decides when compression is triggered).
KC-Bench is introduced, a controlled multi-turn benchmark for measuring model-level behavior across world-knowledge conflicts, input inconsistencies, and multi-source temporal conflicts, and provides a reproducible diagnostic for developing conflict-aware reasoning and execution safeguards.
Yaxing Lyu, Sheng-Jie Zhou, B. Toh et al.· 2 citations
Related blog posts
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 9, 2026
Assistant Professor Benjamin Lindquist studies how engineers deliberately injected randomness into early AI, giving rise to deep-learning models that sometimes hallucinate.