{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.19613", "kind": "paper", "url": "https://arxiv.org/abs/2606.19613"}, {"id": "gh:amazon-science/staminabench", "kind": "repo", "url": "https://github.com/amazon-science/StaminaBench"}], "authors": ["Vlad Sobal", "Shuo Yang", "Yuting Zhang", "Wei Xia", "Stefano Soatto"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.19613", "source_url": "https://arxiv.org/abs/2606.19613"}, "categories": ["Code & Software", "General AI", "cs.SE"], "description": {"en": "StaminaBench stress-tests coding agents over 100 interaction turns of change requests, measuring how many consecutive turns they handle before failing, with programmatically generated tasks and black-box HTTP evaluation."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.19613", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.19613", "title": "StaminaBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.19613", "modality": null, "name": "StaminaBench", "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_staminabench_cb49af23", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "2901d8d2bae6a090f8c72e2f242e076f7c4818d491251dacb72020c2d5900931", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.19613"}, "publisher": null, "released": "2026-06-17", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.19613", "source_url": "https://arxiv.org/abs/2606.19613"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-19613", "source": "claire_radar", "source_benchmark_id": "2606.19613", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Machine Generated", "applicationDomains": ["General AI"], "area": "Code & Software", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/amazon-science/StaminaBench", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.19613", "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", "state": "unavailable"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "d57ee1987a79485a6b3445097afcaa1f466de6d28e5d360fb520bf31eddc373a", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "StaminaBench stress-tests coding agents over 100 interaction turns of change requests, measuring how many consecutive turns they handle before failing, with programmatically generated tasks and black-box HTTP evaluation.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing."}, "familyId": "bmf_5030babf657a", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "linked-from-report", "linkedUrl": "http://github.com/amazon-science/StaminaBench", "url": "https://arxiv.org/abs/2606.19613"}], "scope": "benchmark_repo", "url": "https://github.com/amazon-science/StaminaBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_staminabench_cb49af23", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/amazon-science/StaminaBench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.19613", "paper": "https://arxiv.org/abs/2606.19613", "pdf": "https://arxiv.org/pdf/2606.19613", "project": null, "report": "https://arxiv.org/abs/2606.19613"}, "metrics": [], "motivation": "We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing.", "name": "StaminaBench", "oneLine": "StaminaBench stress-tests coding agents over 100 interaction turns of change requests, measuring how many consecutive turns they handle before failing, with programmatically generated tasks and black-box HTTP evaluation.", "paperTitle": "StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.19613", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-17", "source": {"authors": ["Vlad Sobal", "Shuo Yang", "Yuting Zhang", "Wei Xia", "Stefano Soatto"], "categories": ["cs.SE", "cs.AI"], "id": "2606.19613", "title": "StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns", "type": "arxiv", "url": "https://arxiv.org/abs/2606.19613"}, "sourceUpdatedAt": "2026-06-17T21:36:09Z", "topics": ["cs.SE"], "venueAttempts": [], "whyItMatters": "Fills the gap in evaluating multi-turn coding agent stamina, which is critical for real-world vibe-coding sessions that often extend over many turns."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
