AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

Evolving Generalist Virtual Agents with Generative and Associative Memory

Zhenkui Zhang, Wendong Bu, Kaihang Pan, Bingchen Miao, Wenqiao Zhang, Guoming Wang, Wei Ji, Rui Tang, Juncheng Li, Siliang Tang

AAAI 2026 · Multi-agent systems and games › LLM agents · approach: Agentic benchmarks and orchestration · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a memory architecture for generalist virtual agents that moves beyond isolated experience fragments and brittle semantic retrieval to enable evolving, novel solution synthesis.

Problem

Generalist Virtual Agents built on MLLMs fail to evolve beyond existing trajectories because their memory treats experiences as isolated fragments and relies on brittle semantic retrieval, preventing synthesis of novel solutions from disparate knowledge.

Approach

Introduces generative and associative memory for GVAs so that stored experiences are connected and recombined, rather than retrieved as isolated fragments by semantic similarity. Details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add memory recombination (ablation)CA3Mem · OSWorld (15-step) overallsuccess rate (%): 16.80 -> 23.85Improved
add memory recombination (ablation)CA3Mem · OSWorld (15-step) Workflowsuccess rate (%): 8.91 -> 15.84Improved
add memory recombination (ablation)CA3Mem · OSWorld (15-step) Officesuccess rate (%): 8.55 -> 14.53Improved
add spreading activation retrieval (ablation)CA3Mem · OSWorld (15-step) overallsuccess rate (%): 14.91 -> 23.85Improved
add spreading activation retrieval (ablation)CA3Mem · OSWorld (15-step) Dailysuccess rate (%): 16.67 -> 30.77Improved
add spreading activation retrieval (ablation)CA3Mem · OSWorld (15-step) Workflowsuccess rate (%): 7.92 -> 15.84Improved
apply CA3Mem to Claude-3 backboneClaude-3 · OSWorld (15-step)success rate (%): 4.41 -> 18.43Improved
apply CA3Mem to Qwen2.5-VL backboneQwen2.5-VL · OSWorld (15-step)success rate (%): 8.80 -> 20.05Improved
apply CA3Mem to GPT-4o on Workflow tasksGPT-4o · OSWorld (15-step) Workflowsuccess rate (%): 5.58 -> 15.84Improved
accumulate memory through four stages of synthetic tasksCA3Mem · OSWorld (15-step) Dailysuccess rate (%): 6.41% to 30.07%Improved
use CA3Mem instead of AWMWebArena agent · WebArena (Reddit)success rate (%): 50.9 -> 63.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy