{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.14441", "kind": "paper", "url": "https://arxiv.org/abs/2608.14441"}, {"id": "gh:thunlp/pace-bench", "kind": "repo", "url": "https://github.com/thunlp/PACE-Bench"}], "authors": ["Yuhao Zhan", "Bingxiang He", "Zecong Tang", "Chaojun Xiao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.14441", "source_url": "https://arxiv.org/abs/2608.14441"}, "categories": ["Agents", "General AI", "Language & Knowledge", "Self-Evolution"], "description": {"en": "PACE-Bench is a simulator-grounded benchmark with 144 source-to-target adaptation pairs across six physics domains. Each pair presents a code-driven design that succeeds in a source environment but fails in a mutated target environment, and agents must iteratively adapt the design using diagnostic sandbox feedback within a limited attempt budget."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.14441", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.14441", "title": "PACE-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.14441", "modality": null, "name": "PACE-Bench", "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_pace-bench_0530dc3f", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "35f398dc0260d6738a8de06df4e24b66489578dbf3f598ab2dffad481083bda8", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2608.14441"}, "publisher": null, "released": "2026-08-14", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.14441", "source_url": "https://arxiv.org/abs/2608.14441"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-14441", "source": "claire_radar", "source_benchmark_id": "2608.14441", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/thunlp/PACE-Bench", "githubScope": "benchmark_repo", "githubStars": 6, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 29, "hfPaperUrl": "https://huggingface.co/papers/2608.14441", "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": "2026-09-30", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Interactive Environment", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "190109402f0e67b5f2d47cd58eab25ad418d453350fc095b2733be22d22423bf", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "PACE-Bench is a simulator-grounded benchmark with 144 source-to-target adaptation pairs across six physics domains. Each pair presents a code-driven design that succeeds in a source environment but fails in a mutated target environment, and agents must iteratively adapt the design using diagnostic sandbox feedback within a limited attempt budget.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simulator-grounded benchmark of 144 source-to-target adaptation pairs across six physics domains."}, "familyId": "bmf_91838779e882", "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/thunlp/PACE-Bench", "url": "https://github.com/thunlp/PACE-Bench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/thunlp/PACE-Bench", "url": "https://huggingface.co/papers/2608.14441"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/thunlp/PACE-Bench.", "url": "https://arxiv.org/abs/2608.14441"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/thunlp/PACE-Bench", "url": "https://arxiv.org/abs/2608.14441"}], "scope": "benchmark_repo", "url": "https://github.com/thunlp/PACE-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_pace-bench_0530dc3f", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/thunlp/PACE-Bench", "data": null, "hfPaper": "https://huggingface.co/papers/2608.14441", "paper": "https://arxiv.org/abs/2608.14441", "pdf": "https://arxiv.org/pdf/2608.14441", "project": null, "report": "https://arxiv.org/abs/2608.14441"}, "metrics": [], "motivation": "Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change.", "name": "PACE-Bench", "oneLine": "PACE-Bench is a simulator-grounded benchmark with 144 source-to-target adaptation pairs across six physics domains. Each pair presents a code-driven design that succeeds in a source environment but fails in a mutated target environment, and agents must iteratively adapt the design using diagnostic sandbox feedback within a limited attempt budget.", "paperTitle": "PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.14441", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "THUNLP", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/thunlp/PACE-Bench"}], "ranking": {"90d": {"confidence": "Medium", "coverage": 0.7, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.2597440249823917, "value": 6}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.6009732431336289, "value": 29}}, "confidence": "Medium", "coverage": 0.7, "rank": 154, "score": 38}, "method": "attention-ranking-v13", "rank": 154, "score": 38}}, "readiness": "Runnable", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-08-14", "source": {"authors": ["Yuhao Zhan", "Bingxiang He", "Zecong Tang", "Chaojun Xiao"], "categories": ["cs.AI"], "id": "2608.14441", "title": "PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments", "type": "arxiv", "url": "https://arxiv.org/abs/2608.14441"}, "sourceUpdatedAt": "2026-08-14T16:25:43Z", "topics": ["Self-Evolution", "Agents"], "venueAttempts": [], "whyItMatters": "Existing self-evolving agent evaluations assume fixed execution conditions and do not test recovery after environmental shifts. PACE-Bench provides a repeatable protocol to assess an agent's ability to adapt to changing physics, offering insight into the reliability of different self-evolving methods."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
