{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.22902", "kind": "paper", "url": "https://arxiv.org/abs/2606.22902"}, {"id": "gh:lancezpf/agent-as-a-router", "kind": "repo", "url": "https://github.com/LanceZPF/agent-as-a-router"}], "categories": ["Agents", "Agents & Tool Use", "General AI"], "description": {"en": "CodeRouterBench is an evaluation environment for comparing model routers on streaming coding tasks."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.22902", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.22902", "title": "CodeRouterBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.22902", "modality": null, "name": "CodeRouterBench", "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_coderouterbench_7fa606a9", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "85a70f4143e9e85ac4432e7718c5528535769ff19837bc5e3dddb0fa875a0978", "relation": "evaluates_only", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T20:19:32.953828Z", "source_url": "https://arxiv.org/abs/2606.22902"}, "publisher": null, "released": "2026-06-22", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.22902", "source_url": "https://arxiv.org/abs/2606.22902"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-22902", "source": "claire_radar", "source_benchmark_id": "2606.22902", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Agents & Tool Use", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The benchmark may attract limited near-term interest due to its large task collection and focus on coding-model routing, but attention cannot determine admission.", "score": 3}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T20:19:32.953828Z", "inputHash": "e295f7e14445cb6114b6aafbaa86378754239117f8f5d6c8fbef901b50f1d742", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "introduce CodeRouterBench", "canonicalNameSource": "abstract", "decisionReason": "The source formally names the benchmark, defines a verified-score streaming routing environment, specifies regret-based comparison, and provides a released public repository.", "evaluationEvidence": {"reuse": "Codes and benchmarks are released at https://github.com/LanceZPF/agent-as-a-router.", "scoring": "enabling regret-based router comparison on streaming tasks", "task": "an evaluation environment comprising ~10K task instances with verified scores from 8 frontier LLMs"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T20:19:32.953828Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "CodeRouterBench is an evaluation environment for comparing model routers on streaming coding tasks.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "evaluation protocol evidence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We instantiate this framework as ACRouter, composed of an Orchestrator, a Verifier, a Memory module, and introduce CodeRouterBench, an evaluation environment comprising ~10K task instances with verified scores from 8 frontier LLMs, enabling regret-based router comparison on streaming tasks."}, "familyId": "bmf_a5d24d9eb8d2", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_coderouterbench_7fa606a9", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/LanceZPF/agent-as-a-router", "data": null, "paper": "https://arxiv.org/abs/2606.22902", "pdf": "https://arxiv.org/pdf/2606.22902", "project": null, "report": "https://arxiv.org/abs/2606.22902"}, "metrics": [], "motivation": "Real-world users typically have access to multiple Large Language Models (LLMs) from different providers, and these LLMs often excel at distinct domains, yet none dominate all.", "name": "CodeRouterBench", "oneLine": "CodeRouterBench is an evaluation environment for comparing model routers on streaming coding tasks.", "paperTitle": "Agent-as-a-Router: Agentic Model Routing for Coding Tasks", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.22902", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-06-22", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Pengfei Zhou", "Zhiwei Tang", "Yixing Ma", "Jiasheng Tang", "Yizeng Han", "Zhenglin Wan", "Fanqing Meng", "Wei Wang", "Bohan Zhuang", "Wangbo Zhao", "Yang You"], "categories": ["cs.AI"], "id": "2606.22902", "title": "Agent-as-a-Router: Agentic Model Routing for Coding Tasks", "type": "arxiv", "url": "https://arxiv.org/abs/2606.22902"}, "sourceUpdatedAt": "2026-06-26T11:55:14Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "It targets execution-grounded routing across multiple coding models and includes approximately 10,000 task instances with verified model scores."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
