{"record": {"aliases": ["SWE-InfraBench: Evaluating Language Models on Cloud Infrastructure Code"], "artifacts": [{"id": "arxiv:2606.05249", "kind": "paper", "url": "https://arxiv.org/abs/2606.05249"}], "categories": ["Code & Software", "Software & AI Compute", "cs.SE"], "description": {"en": "A benchmark of incremental infrastructure-as-code modifications in real-world AWS CDK repositories, evaluated through provided test cases."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.05249", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.05249", "title": "SWE-InfraBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.05249", "modality": null, "name": "SWE-InfraBench", "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_swe-infrabench-evaluating-language-models-_0fcf43c6", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "9fecfc142c90b896b324d9430705fd48029920b309d7e6beb324c3dfba15b723", "relation": "evaluates_only", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T20:00:48.581944Z", "source_url": "https://arxiv.org/abs/2606.05249"}, "publisher": null, "released": "2026-06-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.05249", "source_url": "https://arxiv.org/abs/2606.05249"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-05249", "source": "claire_radar", "source_benchmark_id": "2606.05249", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["SWE-InfraBench: Evaluating Language Models on Cloud Infrastructure Code"], "annotation": "Unknown", "applicationDomains": ["Software & AI Compute"], "area": "Code & Software", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The benchmark addresses a specialized but practically relevant software-engineering domain and provides reusable data.", "score": 3}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T20:00:48.581944Z", "inputHash": "2b7525e48c00f52ee300bb37fffd84b3535e3d74b4dcd538ba63ce6ccafb9d0d", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "SWE-InfraBench: Evaluating Language Models on Cloud Infrastructure Code", "canonicalNameSource": "paper_title", "decisionReason": "Publish: the source formally names the benchmark, specifies a repeatable code-modification task, defines test-case passage as the success criterion, and supplies a currently available public dataset link.", "evaluationEvidence": {"reuse": "The SWE-InfraBench dataset is available at: https://www.kaggle.com/datasets/64e59070fd51c0278560b01eb5dc4f3c447d5268cdabe5a350d2969e4413fea5", "scoring": "success determined by passing provided test cases", "task": "Each example requires models to implement changes to existing codebases based on natural language instructions"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T20:00:48.581944Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark of incremental infrastructure-as-code modifications in real-world AWS CDK repositories, evaluated through provided test cases.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We present SWE-InfraBench, a diverse evaluation dataset sourced from dozens of real-world IaC codebases that challenge LLMs to perform realistic code modifications in AWS CDK repositories."}, "familyId": "bmf_7b9238b38c5b", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_swe-infrabench-evaluating-language-models-_0fcf43c6", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Software & Cloud"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2606.05249", "pdf": "https://arxiv.org/pdf/2606.05249", "project": "https://www.kaggle.com/datasets/64e59070fd51c0278560b01eb5dc4f3c447d5268cdabe5a350d2969e4413fea5", "report": "https://arxiv.org/abs/2606.05249"}, "metrics": [], "motivation": "Building infrastructure-as-code (IaC) in cloud computing is a critical task, underpinning the reliability, scalability, and security of modern software systems.", "name": "SWE-InfraBench", "oneLine": "A benchmark of incremental infrastructure-as-code modifications in real-world AWS CDK repositories, evaluated through provided test cases.", "paperTitle": "SWE-InfraBench: Evaluating Language Models on Cloud Infrastructure Code", "primaryDomain": "Software & AI Compute", "publication": {"evidence": "Accepted to NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2606.05249", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle", "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-06-03", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Natalia Tarasova", "Enrique Balp-Straffon", "Aleksei Iancheruk", "Yevhenii Sielskyi", "Nikita Kozodoi", "Liam H. Byrne", "Jack Butler", "Dayuan Jiang", "Marcin Czelej", "Andrew Ang", "Yash Shah", "Roi Blanco", "and Sergei Ivanov"], "categories": ["cs.SE"], "id": "2606.05249", "title": "SWE-InfraBench: Evaluating Language Models on Cloud Infrastructure Code", "type": "arxiv", "url": "https://arxiv.org/abs/2606.05249"}, "sourceUpdatedAt": "2026-06-03T12:48:53Z", "topics": ["cs.SE"], "venueAttempts": [{"decisionRaw": "Accepted to NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle", "evidence": [{"level": "author-claim", "observedAt": "2026-08-19T11:05:13.395391Z", "rawValue": "Accepted to NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2606.05249"}], "reviewStatus": "accepted", "venueName": "NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle"}], "whyItMatters": "It extends software-engineering evaluation to realistic cloud-infrastructure maintenance tasks involving resource dependencies and existing codebases."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
