Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World

Yusong Lin, Xinyuan Liang, Haiyang Wang, Qipeng Gu, Siqi Cheng, Jiangui Chen, Shuzhe Wu, Feiyang Pan, Lue Fan, Sanyuan Zhao, Dandan Tu

arXiv:2605.26086 · 2026-05-26 공개 · arXiv · PDF

long-horizon agent-evaluation contextual-reasoning proactive-assistance noise-robustness multi-device gui-cli data-generation

Abstract

Large language model agents are increasingly envisioned as always-on personal assistants with access to anything relevant in the user's digital world. Yet current systems operate over only narrow slices of that world, limiting context-sensitive reasoning and effective assistance. Existing benchmarks similarly provide only partial user state and therefore fail to capture performance in such a broad, always-on setting. To address this gap, we introduce Claw-Anything, a benchmark that expands agent context along three dimensions: long-horizon activity histories, interdependent backend services, and integrated GUI and CLI interaction across multiple devices. To instantiate this setting, we simulate months of user activity through multi-round event injection, producing complex world states and realistic noise, including irrelevant events and conflicting signals. Agents must reason over rich contextual environments while remaining robust to such noise. This expanded scope also enables the evaluation of proactive assistance, requiring agents to anticipate user needs and deliver timely recommendations. Experiments show that GPT-5.5 achieves only 34.5% pass@1, substantially below prior benchmarks, underscoring a gap between current agent capabilities and the demands of always-on personal assistance. Alongside the benchmark, we release an automated data-generation pipeline that yields 2,000 training environments and improves the base model by 23.7%, demonstrating its utility of scalable data infrastructure.

한국어 요약

📋 한 줄 요약

**[Personal Assistant 벤치 / Always-On Agents]** Claw-Anything이 long-horizon activity·interdependent backend service·multi-device GUI+CLI를 결합한 always-on assistant 벤치 제시, GPT-5.5도 pass@1 34.5%, 2,000 학습 환경으로 base 모델 23.7% 개선.

🎯 핵심 기여도

💡 핵심 아이디어

실제 always-on personal assistant 평가는 단일 slice가 아닌 long-horizon activity·interdependent service·multi-device GUI+CLI를 모두 포함하고 irrelevant·conflicting noise까지 주입하는 환경이 필요하며, 이 broad scope가 현재 frontier agent의 큰 capability gap을 노출한다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: Always-on personal assistant 평가의 새 표준 정립, 3차원 context 확장으로 실세계에 더 가까운 벤치, GPT-5.5조차 34.5%로 frontier capability gap 노출, 학습 환경 2,000개와 23.7% 개선의 추가 가치. **한계**: 시뮬레이션 충실도와 실 사용자 데이터의 격차, 200 시나리오 수준의 cognitive profile 다양성 한계, irrelevant·conflicting noise의 분포 설계 의존성.

🚀 실용적 활용