2026· Annual Meeting of the Association for Computational Linguistics· pp. 30424-30451· 2 citations· ⚡ 1 influential· 39 references
Computer Science
TL;DR
MUSE is a framework that enables iterative self-improvement through a hierarchical Memory Module that organizes cross-domain insights to facilitate the orchestration of long-horizon workflows and demonstrates that MUSE’s performance scales with the accumulation of insights and exhibits strong cross-task transferability.
Abstract
Large Language Models often struggle with complex, multi-step operational tasks because they remain static during inference and cannot learn from past experience. To address this, we propose MUSE, a framework that enables iterative self-improvement through a hierarchical Memory Module. MUSE organizes cross-domain insights to facilitate the orchestration of long-horizon workflows. The core of our approach is an autonomous post-execution critique mechanism: after completing each sub-task, the system analyzes its operational logs and distills raw execution data into structured, reusable knowledge. This allows the agent to evolve dynamically rather than relying on fixed parameters. Evaluated on the rigorous TAC productivity benchmark, MUSE achieves new state-of-the-art results, significantly out-performing previous methods using only the streamlined Gemini-2.5 Flash model. Our analysis demonstrates that MUSE’s performance scales with the accumulation of insights and exhibits strong cross-task transferability, marking a key step toward autonomous systems capable of lifelong learning in professional environments. Demo videos can be found in our supplementary materials.
Recent advances in large language models have enabled a new class of agentic data science systems that allow users to complete complex data science workflows through natural language. Although these systems can significantly reduce manual effort, it remains difficult to diagnose their behavior and steer the reasoning process when failures or unexpected outputs occur. We present MUSE, an interactive meta-agent that enhances user understanding and control of agentic data science systems by (1) dynamically restructuring low-level execution traces into multiple semantic levels that support navigation from high-level overviews to low-level implementation details; (2) enabling users to reference specific workflow steps in context to ask grounded questions, provide feedback, and revise problematic steps without manually locating relevant execution history; and (3) supporting mixed-initiative steering by surfacing suspicious steps for inspection, scaffolding the repair process, and translating user repair intent into contextualized instructions for the underlying agent. In a between-subjects study (n = 15), MUSE improved task efficiency and increased users'confidence in understanding and steering agentic data science workflows.
Wei-Hao Chen, Weixi Tong, Yuan Tian et al.· 0 citations
Extensive experiments show that ExpG brings consistent improvements across the tool selection, tool calling, and response generation tasks, enabling smaller agents to outperform larger ones that do not use ExpG, suggesting a promising path toward more robust tool use.
Comparisons against stronger model and coding-agent competitors further indicate that both domain-specific agent runtime structure and foundation-model strength matter for autonomous data analysis.
but struggle when confronted with multi-step, complex task planning that requires interaction with external environments. This paper investigates the architecture, implementation, and efficacy of tool-augmented language agents enhanced with iterative self-critique mechanisms. By integrating external application programming interfaces, structured databases, and computational engines, these agents transcend isolated text generation, evolving into active systems capable of executing concrete actions. However, naive tool utilization often results in cascading errors during prolonged execution trajectories. To mitigate this, we introduce an iterative self-critique framework where the agent continuously evaluates its own outputs, identifies logical fallacies or execution failures, and dynamically recalibrates its plan. This research details a comprehensive methodological framework, formalizing the probabilistic decision-making and critique generation processes. Empirical evaluations across simulated complex environments demonstrate that the proposed architecture significantly improves task success rates, minimizes superfluous tool invocations, and enhances error recovery. The findings indicate that integrating reflective cognition paradigms with modular toolsets is essential for deploying autonomous language agents in high-stakes, real-world applications.
Mabel Kwok· International journal of inf...· 0 citations
Learning from experience is critical for developing capable, self-improving large language model (LLM) agents. Existing methods typically extract knowledge from accumulated trajectories via reflection, memory, rules, or skills. However, agents in realistic environments continuously encounter novel tasks, often offering only a one-shot opportunity to improve. These executions yield rich but highly noisy contexts, entangling broadly useful lessons with task-specific artifacts. Critically, prior works rarely validate their effectiveness on complex real-world tasks or isolate the underlying drivers of improvement. To address these gaps, we formulate online harness learning, where a frozen agent improves by continually updating a structured harness across sequential tasks. This formulation enables a systematic study of key self-improvement factors through our proposed Evo-Harness. At its core, context-to-harness skill compilation distills noisy, single-shot executions into reusable skill harnesses for cross-domain and topic-level adaptation. To demonstrate the efficacy of one-shot skill compilation, we evaluate across five realistic benchmarks (TerminalBench2, SWE-bench, CL-Bench, -bench, WebArena-Infinity). Our extensive analysis demonstrates the effectiveness of Evo-Harness and provides a principled understanding of how LLM agents can effectively learn on the fly. Our code is available at https://github.com/A-EVO-Lab/a-evolve/tree/release/evo-harness.
Tianxin Wei, Zhan Shi, Minhua Lin et al.· 2 citations