{"record": {"aliases": ["Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution"], "artifacts": [{"id": "arxiv:2607.13034", "kind": "paper", "url": "https://arxiv.org/abs/2607.13034"}, {"id": "gh:eejyin/do-ai-agents-know-when-a-task-is-simple-toward-complexity-aware-reasoning-and-execution", "kind": "repo", "url": "https://github.com/eejyin/Do-AI-Agents-Know-When-a-Task-Is-Simple-Toward-Complexity-Aware-Reasoning-and-Execution"}], "authors": ["Junjie Yin", "Xinyu Feng"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.13034", "source_url": "https://arxiv.org/abs/2607.13034"}, "categories": ["General AI", "Language & Knowledge", "Reasoning"], "description": {"en": "MSE-Bench measures agent performance and execution redundancy across deterministic editing tasks in a controlled simulator."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.13034", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.13034", "title": "MSE-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Microsoft (United States)", "University of Tennessee at Knoxville"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.13034", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.13034", "source_url": "https://openalex.org/W7168428854"}, "key": "claire-radar:2607.13034", "modality": null, "name": "MSE-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_do-ai-agents-know-when-a-task-is-simple-to_24b60640", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "a942325d2fa15ee05e56e675aafec02027a9dd6b4f577a9d1855ef2b953e4bee", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T17:27:47.088316Z", "source_url": "https://arxiv.org/abs/2607.13034"}, "publisher": null, "released": "2026-07-14", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.13034", "source_url": "https://arxiv.org/abs/2607.13034"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-13034", "source": "claire_radar", "source_benchmark_id": "2607.13034", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Agent efficiency and cost-aware execution are timely topics, and the deterministic released benchmark offers a concrete comparison target.", "score": 4}, "capabilities": ["Reasoning"], "confidence": "Medium", "construction": "Interactive Environment", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T17:27:47.088316Z", "inputHash": "9280ae9a8e7fdf87a2a08451774198fc48e6d2f2d488f92876ef4aed63dbe2b3", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "MSE-Bench--a deterministic benchmark of 121 edits in a capability-controlled simulator", "canonicalNameSource": "abstract", "decisionReason": "The abstract formally names a deterministic benchmark, defines its edit-task suite and comparative measures, and the comments provide a public benchmark repository.", "evaluationEvidence": {"reuse": "Code and benchmark: https://github.com/eejyin/Do-AI-Agents-Know-When-a-Task-Is-Simple-Toward-Complexity-Aware-Reasoning-and-Execution", "scoring": "100% success while cutting cost by 85%, tokens by 91%, and inspected files by 92%", "task": "a deterministic benchmark of 121 edits in a capability-controlled simulator"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T17:27:47.088316Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MSE-Bench measures agent performance and execution redundancy across deterministic editing tasks in a controlled simulator.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We release the framework and benchmark."}, "familyId": "bmf_b226547faac6", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_do-ai-agents-know-when-a-task-is-simple-to_24b60640", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/eejyin/Do-AI-Agents-Know-When-a-Task-Is-Simple-Toward-Complexity-Aware-Reasoning-and-Execution", "data": null, "paper": "https://arxiv.org/abs/2607.13034", "pdf": "https://arxiv.org/pdf/2607.13034", "project": null, "report": "https://arxiv.org/abs/2607.13034"}, "metrics": [], "motivation": "Large language model (LLM) agents increasingly automate multi-step engineering and informatics workflows, yet they rarely ask how much effort a task actually requires.", "name": "MSE-Bench", "oneLine": "MSE-Bench measures agent performance and execution redundancy across deterministic editing tasks in a controlled simulator.", "paperTitle": "Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.13034", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-07-14", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Junjie Yin and Xinyu Feng"], "categories": ["cs.AI", "cs.CL", "cs.SE", "cs.SY", "eess.SY"], "id": "2607.13034", "title": "Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution", "type": "arxiv", "url": "https://arxiv.org/abs/2607.13034"}, "sourceUpdatedAt": "2026-07-14T17:59:31Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "It supports comparisons of whether agents can complete edits successfully while limiting unnecessary cost, token use, and file inspection."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
