{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.31174", "kind": "paper", "url": "https://arxiv.org/abs/2606.31174"}, {"id": "gh:aiming-lab/clawarena", "kind": "repo", "url": "https://github.com/aiming-lab/ClawArena"}, {"kind": "website", "url": "https://www.clawarena.cc/"}], "authors": ["Kaiwen Xiong", "Haonian Ji", "Shi Qiu", "Zeyu Zheng", "Cihang Xie", "Xinyu Ye", "Huaxiu Yao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.31174", "source_url": "https://arxiv.org/abs/2606.31174"}, "categories": ["General AI", "Multimodal"], "description": {"en": "ClawArena-Team evaluates a single text-only LLM's ability to manage a fixed, locally served pool of subagents (LLM, VLM, omni) across 41 multi-turn, multimodal, multi-directory scenarios with 258 evaluation rounds and 72 staged updates. Scoring is execution-based via shell commands, producing a Subagent-Management Score (SMS) that multiplies task correctness by a least-privilege and modality-routing factor, without LLM judges."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.31174", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.31174", "title": "ClawArena-Team"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.31174", "modality": null, "name": "ClawArena-Team", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_clawarena-team_8c93773c", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "a43a0e7818e40b23d4c897ed2aac7798865981b021923675c27e7fea77f0f90c", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.31174"}, "publisher": null, "released": "2026-06-30", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.31174", "source_url": "https://arxiv.org/abs/2606.31174"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-31174", "source": "claire_radar", "source_benchmark_id": "2606.31174", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/aiming-lab/ClawArena", "githubScope": "benchmark_repo", "githubStars": 69, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2606.31174", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-10-04", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "9666a2ac4bc96807405ca733dd9d6cc3d5253f3cac5e46e4dde649dd478f78fc", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ClawArena-Team evaluates a single text-only LLM's ability to manage a fixed, locally served pool of subagents (LLM, VLM, omni) across 41 multi-turn, multimodal, multi-directory scenarios with 258 evaluation rounds and 72 staged updates. Scoring is execution-based via shell commands, producing a Subagent-Management Score (SMS) that multiplies task correctness by a least-privilege and modality-routing factor, without LLM judges.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce ClawArena-Team, a benchmark of 41 multi-turn, multimodal, multi-directory scenarios spanning 258 evaluation rounds and 72 staged updates that measures this management ability."}, "familyId": "bmf_c7eb98430ee5", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/aiming-lab/ClawArena", "url": "https://github.com/aiming-lab/ClawArena"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/aiming-lab/ClawArena", "url": "https://huggingface.co/papers/2606.31174"}, {"basis": "linked-from-project", "linkedUrl": "https://github.com/aiming-lab/ClawArena", "url": "https://www.clawarena.cc/"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/aiming-lab/ClawArena", "url": "https://arxiv.org/abs/2606.31174"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/aiming-lab/ClawArena.", "url": "https://arxiv.org/abs/2606.31174"}], "scope": "benchmark_repo", "url": "https://github.com/aiming-lab/ClawArena", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_clawarena-team_8c93773c", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/aiming-lab/ClawArena", "data": null, "hfPaper": "https://huggingface.co/papers/2606.31174", "paper": "https://arxiv.org/abs/2606.31174", "pdf": "https://arxiv.org/pdf/2606.31174", "project": "https://www.clawarena.cc/", "report": "https://arxiv.org/abs/2606.31174"}, "metrics": [], "motivation": "Production large language-model (LLM) agents are increasingly deployed not as lone problem-solvers but as managers: a main model creates specialized subagents, delegates work, and orchestrates their parallel, asynchronous returns through dynamic workflows.", "name": "ClawArena-Team", "oneLine": "ClawArena-Team evaluates a single text-only LLM's ability to manage a fixed, locally served pool of subagents (LLM, VLM, omni) across 41 multi-turn, multimodal, multi-directory scenarios with 258 evaluation rounds and 72 staged updates. Scoring is execution-based via shell commands, producing a Subagent-Management Score (SMS) that multiplies task correctness by a least-privilege and modality-routing factor, without LLM judges.", "paperTitle": "ClawArena-Team: Benchmarking Subagent Orchestration and Dynamic Workflows in Language-Model Agents", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.31174", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-30", "source": {"authors": ["Kaiwen Xiong", "Haonian Ji", "Shi Qiu", "Zeyu Zheng", "Cihang Xie", "Xinyu Ye", "Huaxiu Yao"], "categories": ["cs.AI"], "id": "2606.31174", "title": "ClawArena-Team: Benchmarking Subagent Orchestration and Dynamic Workflows in Language-Model Agents", "type": "arxiv", "url": "https://arxiv.org/abs/2606.31174"}, "sourceUpdatedAt": "2026-07-02T07:39:52Z", "topics": ["Multimodal"], "venueAttempts": [], "whyItMatters": "Existing agent benchmarks measure a policy's own task-solving or emergent behavior of fixed multi-agent systems, but not the leadership capability of a single model orchestrating subagents. ClawArena-Team fills this gap by isolating management skill from raw capability, supporting decisions on model selection for delegation-heavy workflows and revealing cost-quality trade-offs."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
