Research map / Multi-agent systems and games / LLM agents
HippoCamp: Benchmarking Contextual Agents on Personal Computers
Zhe YANG, Shulin Tian, Kairui Hu, Shuai Liu, Hoang-Nhat Nguyen, Yichi Zhang, Zujin Guo, Mengying Yu, Zinan Zhang, Jingkang Yang, Chen Change Loy, Ziwei Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use iterative search agent instead of single-pass RAG | Standard RAG vs ReAct (Qwen3) · HippoCamp factual retention (overall) | F1 (%): 30.0 -> 43.1 | Improved |
| use iterative search agent instead of single-pass RAG | Standard RAG vs ReAct (Qwen3) · HippoCamp profiling (overall) | accuracy (%): 26.7 -> 13.5 | Hurt |
| use autonomous ChatGPT Agent Mode instead of search agents | ReAct (Qwen3) · HippoCamp profiling (overall) | accuracy (%): 13.5 -> 48.3 | Improved |
| profiling vs factual retention task (Search-R1) | Search-R1 · HippoCamp overall | accuracy (%): 25.3 -> 5.0 | Hurt |
| profiling vs factual retention task (ReAct Qwen3) | ReAct (Qwen3-30B-A3B) · HippoCamp overall | accuracy (%): 28.5 -> 13.5 | Hurt |
| move from legal Adam profile to media-rich Bei profile | ChatGPT Agent Mode · HippoCamp factual retention | accuracy (%): 90.3 -> 31.2 | Hurt |
| high search F1 but weak perception | ChatGPT Agent Mode · HippoCamp profiling | accuracy (%): search 56.5 vs perception 28.5 | Mixed |
| higher retrieval F1 does not give higher answer accuracy | ReAct (Qwen3-30B-A3B) · HippoCamp factual retention | accuracy (%): 43.1 F1 vs 28.5 Acc | Mixed |
| correct answers without grounding evidence | Terminal Agent (GPT-5.2) · HippoCamp factual retention | accuracy (%): 24.6 F1 vs 48.2 Acc | Mixed |
Closest papers by meaning
- SMMBench: A Benchmark for Source-Distributed Multimodal Agent Memory (NeurIPS 2026 · similarity 0.87)
- AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications (ICML 2026 · similarity 0.87)
- Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependencies (NeurIPS 2026 · similarity 0.86)
- Embodied Agents Meet Personalization: Investigating Challenges and Solutions Through the Lens of Memory Utilization (ICLR 2026 · similarity 0.86)
- OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments (NeurIPS 2024 · similarity 0.86)
- Ego2Web: A Web Agent Benchmark Grounded in Egocentric Videos (CVPR 2026 · similarity 0.86)
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.
