Managing Procedural Memory in LLM Agents: Control, Adaptation, and Evaluation

Hugging Face Daily Papers Papers

Summary

Introduces AFTER, a benchmark of 382 enterprise tasks to evaluate procedural memory in LLM agents, showing that skill transfer improves performance across tasks, roles, and model backbones, with some skills generalizing broadly while others specialize.

Procedural memory is increasingly used to improve LLM agents on recurring workplace tasks, yet its ability to produce reusable skills remains poorly understood. We introduce AFTER, a benchmark of 382 realistic enterprise tasks spanning six professional roles and 22 procedural skills, designed to evaluate how skills transfer across tasks, roles, and model backbones. The benchmark includes controlled evaluation settings for local improvement, cross-task transfer, cross-role transfer, and cross-model generalization. Experiments show that procedural memory delivers consistent gains in industrial workflows: a single refinement round improves aggregate performance by 3.7-6.7 points, while skills evolved from diverse multi-model execution traces achieve 73.1% cross-model test accuracy, outperforming all single-model trace sources. We further find that some skills generalize broadly across tasks and models, whereas others become specialized to role-specific workflows and lose effectiveness under transfer. These results provide practical guidance for building, evaluating, and deploying procedural memory systems in production agent platforms.
Original Article
View Cached Full Text

Cached at: 07/01/26, 11:42 AM

Paper page - Managing Procedural Memory in LLM Agents: Control, Adaptation, and Evaluation

Source: https://huggingface.co/papers/2606.23127

Abstract

Procedural memory enhances LLM agents on workplace tasks through skill transfer across roles and models, with varying generalization capabilities affecting deployment strategies.

Procedural memoryis increasingly used to improveLLM agentson recurring workplace tasks, yet its ability to produce reusable skills remains poorly understood. We introduce AFTER, a benchmark of 382 realisticenterprise tasksspanning six professional roles and 22procedural skills, designed to evaluate how skills transfer across tasks, roles, and model backbones. The benchmark includes controlled evaluation settings for local improvement,cross-task transfer,cross-role transfer, andcross-model generalization. Experiments show thatprocedural memorydelivers consistent gains in industrial workflows: a single refinement round improvesaggregate performanceby 3.7-6.7 points, while skills evolved from diversemulti-model execution tracesachieve 73.1% cross-model test accuracy, outperforming all single-model trace sources. We further find that some skills generalize broadly across tasks and models, whereas others become specialized to role-specific workflows and lose effectiveness under transfer. These results provide practical guidance for building, evaluating, and deployingprocedural memorysystems in production agent platforms.

View arXiv pageView PDFGitHub1Add to collection

Get this paper in your agent:

hf papers read 2606\.23127

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.23127 in a model README.md to link it from this page.

Datasets citing this paper1

#### DavydenkoGr/AFTER Updated1 day ago • 1.01k • 2

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.23127 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

AdMem: Advanced Memory for Task-solving Agents

arXiv cs.AI

This paper introduces AdMem, a unified memory framework for LLM-based agents that integrates semantic, episodic, and procedural memory with a bi-level short-term and long-term store, using a multi-agent architecture for automatic memory generation and adaptive retrieval. Experiments show improved robustness and success on long multi-turn tasks.

From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanisms

Hugging Face Daily Papers

This survey paper proposes an evolutionary framework for LLM agent memory mechanisms, categorizing their development into three stages: storage, reflection, and experience. It analyzes core drivers such as long-range consistency and continual learning to provide design principles for next-generation agents.

MemGym: a Long-Horizon Memory Environment for LLM Agents

arXiv cs.CL

MemGym is a benchmark for evaluating memory formation in LLM agents over long-horizon tasks, unifying existing agent gyms and synthetic pipelines with memory-isolated scores. It spans tool-use dialogue, multi-turn search, coding, and computer use, and includes a lightweight reward model (MemRM) for efficient evaluation.