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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add memory recombination (ablation) | CA3Mem · OSWorld (15-step) overall | success rate (%): 16.80 -> 23.85 | Improved |
| add memory recombination (ablation) | CA3Mem · OSWorld (15-step) Workflow | success rate (%): 8.91 -> 15.84 | Improved |
| add memory recombination (ablation) | CA3Mem · OSWorld (15-step) Office | success rate (%): 8.55 -> 14.53 | Improved |
| add spreading activation retrieval (ablation) | CA3Mem · OSWorld (15-step) overall | success rate (%): 14.91 -> 23.85 | Improved |
| add spreading activation retrieval (ablation) | CA3Mem · OSWorld (15-step) Daily | success rate (%): 16.67 -> 30.77 | Improved |
| add spreading activation retrieval (ablation) | CA3Mem · OSWorld (15-step) Workflow | success rate (%): 7.92 -> 15.84 | Improved |
| apply CA3Mem to Claude-3 backbone | Claude-3 · OSWorld (15-step) | success rate (%): 4.41 -> 18.43 | Improved |
| apply CA3Mem to Qwen2.5-VL backbone | Qwen2.5-VL · OSWorld (15-step) | success rate (%): 8.80 -> 20.05 | Improved |
| apply CA3Mem to GPT-4o on Workflow tasks | GPT-4o · OSWorld (15-step) Workflow | success rate (%): 5.58 -> 15.84 | Improved |
| accumulate memory through four stages of synthetic tasks | CA3Mem · OSWorld (15-step) Daily | success rate (%): 6.41% to 30.07% | Improved |
| use CA3Mem instead of AWM | WebArena agent · WebArena (Reddit) | success rate (%): 50.9 -> 63.2 | Improved |
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.
- Agent S — Agent S: An Open Agentic Framework that Uses Computers Like a Human (ICLR 2025) · better on 3 of 3 shared results
- CRADLE — Cradle: Empowering Foundation Agents towards General Computer Control (ICML 2025) · better on 1 of 1 shared result
- AGUVIS-72B — Aguvis: Unified Pure Vision Agents for Autonomous GUI Interaction (ICML 2025) · better on 1 of 1 shared result
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
- MemEvolve: Meta-Evolution of Agent Memory Systems (ICML 2026 · similarity 0.90)
- G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems (NeurIPS 2025 · similarity 0.89)
- MemGen: Weaving Generative Latent Memory for Self-Evolving Agents (ICLR 2026 · similarity 0.88)
- A-Mem: Agentic Memory for LLM Agents (NeurIPS 2025 · similarity 0.88)
- Optimus-1: Hybrid Multimodal Memory Empowered Agents Excel in Long-Horizon Tasks (NeurIPS 2024 · similarity 0.87)
- AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications (ICML 2026 · similarity 0.87)
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.
