DuoMem: Towards Capable On-Device Memory Agents via Dual-Space Distillation

Hugging Face Daily Papers Papers

Summary

DuoMem is a dual-space distillation framework that transfers procedural problem-solving from large language models to compact student models via context-space and parameter-space distillation, achieving high performance with minimal additional parameters and improved inference speed. It boosts a 4B model from 4.3% to 77.9% task success rate on ALFWorld.

Large Language Model (LLM)-based agents can solve complex procedural tasks by interacting with environments over multiple turns, but this ability typically depends on large models, long contexts, and repeated inference calls. This makes advanced memory-augmented agents difficult to deploy on resource-constrained devices. We introduce DuoMem, a dual-space distillation framework that transfers procedural problem-solving ability from a large teacher model to compact student models. DuoMem distils in two complementary spaces: (1)context-space distillation, which replaces student-generated memories with higher-quality teacher-generated procedural memories prepended to the student's input, and (2)parameter-space distillation, which fine-tunes lightweight LoRA adapters on successful teacher trajectories. Evaluated on ALFWorld, a challenging embodied decision-making benchmark, DuoMem boosts a 4B-parameter model from 4.3% to 77.9% task success rate, closing most of the gap to a 72B teacher model (87.1%), while adding fewer than 10M trainable parameters and only a few megabytes of pre-computed teacher memories. Moreover, the DuoMem-enhanced 4B model completes tasks over 3x faster than the 72B teacher in wall-clock time, making it viable for real-time edge deployment, which would be challenging for the teacher.Extensive ablations across eight models spanning 2B-72B parameters reveal that both distillation axes contribute complementary
Original Article
View Cached Full Text

Cached at: 07/03/26, 03:55 PM

Paper page - DuoMem: Towards Capable On-Device Memory Agents via Dual-Space Distillation

Source: https://huggingface.co/papers/2606.29961

Abstract

DuoMem is a dual-space distillation framework that transfers procedural problem-solving from large language models to compact student models through context-space and parameter-space distillation, achieving high performance with minimal additional parameters and improved inference speed.

Large Language Model(LLM)-based agents can solve complexprocedural tasksby interacting with environments over multiple turns, but this ability typically depends on large models, long contexts, and repeated inference calls. This makes advancedmemory-augmented agentsdifficult to deploy onresource-constrained devices. We introduce DuoMem, adual-space distillationframework that transfers procedural problem-solving ability from a large teacher model to compact student models. DuoMem distils in two complementary spaces: (1)context-space distillation, which replaces student-generated memories with higher-quality teacher-generated procedural memories prepended to the student’s input, and (2)parameter-space distillation, which fine-tunes lightweightLoRA adapterson successful teacher trajectories. Evaluated onALFWorld, a challengingembodied decision-makingbenchmark, DuoMem boosts a 4B-parameter model from 4.3% to 77.9%task success rate, closing most of the gap to a 72B teacher model (87.1%), while adding fewer than 10M trainable parameters and only a few megabytes of pre-computed teacher memories. Moreover, the DuoMem-enhanced 4B model completes tasks over 3x faster than the 72B teacher in wall-clock time, making it viable for real-time edge deployment, which would be challenging for the teacher.Extensive ablations across eight models spanning 2B-72B parameters reveal that both distillation axes contribute complementary

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2606\.29961

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/2606.29961 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.29961 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.29961 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

@omarsar0: // AutoMem // I quite like this idea of metamemory. (bookmark it) This new research from Stanford treats agent's memory…

X AI KOLs Timeline

This Stanford research paper introduces AutoMem, a framework that treats agent memory management as a trainable skill. By optimizing memory structure and proficiency separately, AutoMem improves base agent performance 2x-4x on long-horizon tasks, enabling a 32B open-weight model to compete with frontier systems like Claude Opus 4.5 and Gemini 3.1 Pro Thinking.

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.

SimpleMem: Efficient Lifelong Memory for LLM Agents

Papers with Code Trending

Introduces SimpleMem, an efficient memory framework for LLM agents that uses semantic lossless compression to improve accuracy and reduce token consumption, achieving 26.4% F1 improvement and up to 30x reduction in inference-time token usage.