{"record": {"aliases": ["VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies"], "artifacts": [{"id": "arxiv:2608.12282", "kind": "paper", "url": "https://arxiv.org/abs/2608.12282"}, {"id": "gh:ibm/vakra", "kind": "repo", "url": "https://github.com/IBM/VAKRA"}, {"id": "hf:ibm-research/vakra", "kind": "dataset", "url": "https://huggingface.co/datasets/ibm-research/VAKRA"}], "authors": ["Ankita Rajaram Naik", "Anupama Murthi", "Benjamin Elder", "Siyu Huo", "Raavi Gupta", "Abhinav Jain", "Praveen Venkateswaran", "Abdulhamid Adebayo", "Danish Contractor"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.12282", "source_url": "https://arxiv.org/abs/2608.12282"}, "categories": ["General AI", "Information retrieval", "Language & Knowledge", "Reasoning"], "description": {"en": "Evaluates multi-hop tool-use agents across live executable APIs and retrieval, scoring via policy adherence, exact match, and groundedness judges."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.12282", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.12282", "title": "VAKRA"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["IBM (United States)"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.12282", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.12282", "source_url": "https://openalex.org/W7202344962"}, "key": "claire-radar:2608.12282", "modality": null, "name": "VAKRA", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_vakra-evaluating-multi-hop-reasoning-acros_7a94433a", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "bdb147505e930663b0af3091ec52519100ef7661514f4b0fb4f2c248f0de34ea", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-08-27T04:14:52.836173Z", "source_url": "https://arxiv.org/abs/2608.12282"}, "publisher": null, "released": "2026-08-12", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.12282", "source_url": "https://arxiv.org/abs/2608.12282"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-12282", "source": "claire_radar", "source_benchmark_id": "2608.12282", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/IBM/VAKRA", "githubScope": "benchmark_repo", "githubStars": 71, "hfDailySubmittedAt": null, "hfDataset": "ibm-research/VAKRA", "hfDatasetDownloads": 1530, "hfDatasetLikes": 46, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.12282", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}, "hfDatasetLikes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "VAKRA combines over 8,000 APIs, multi-hop reasoning, and policy constraints with a live leaderboard, broadening its appeal to agent evaluation researchers.", "score": 74}, "capabilities": ["Reasoning", "Information retrieval"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:14:52.836173Z", "inputHash": "d36f88872a27bbf0d65acc7409d419ff1f13373b2e033dd04584754444c43c67", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies", "canonicalNameSource": "paper_title", "decisionReason": "VAKRA is formally named, has a leaderboard, code, and dataset links indicating an ongoing submission-based benchmark.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:14:52.836173Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates multi-hop tool-use agents across live executable APIs and retrieval, scoring via policy adherence, exact match, and groundedness judges.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce VAKRA (e\\textbf{V}aluating \\textbf{A}PI and \\textbf{K}nowledge \\textbf{R}etrieval \\textbf{A}gents), a benchmark of over $8{,}000$ executable APIs across $62$ domains with tasks spanning three settings of increasing difficulty: diverse API interaction styles, multi-hop reasoning over structured APIs, and multi-source reasoning with natural-language tool-use policy constraints."}, "familyId": "bmf_c97fe73eec92", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_vakra-evaluating-multi-hop-reasoning-acros_7a94433a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/IBM/VAKRA", "data": "https://huggingface.co/datasets/ibm-research/VAKRA", "paper": "https://arxiv.org/abs/2608.12282", "pdf": "https://arxiv.org/pdf/2608.12282", "project": null, "report": "https://arxiv.org/abs/2608.12282"}, "metrics": [], "motivation": "Agents deployed in enterprise settings must reason across structured APIs and document collections, yet existing benchmarks evaluate these capabilities in isolation.", "name": "VAKRA", "oneLine": "Evaluates multi-hop tool-use agents across live executable APIs and retrieval, scoring via policy adherence, exact match, and groundedness judges.", "paperTitle": "VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.12282", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "IBM Research", "organizationType": "company-research-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/IBM/VAKRA"}], "ranking": {"90d": {"confidence": "High", "coverage": 0.85, "datasetDownloadRank": 13, "datasetRankPopulation": 85, "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": 0.570858049020576, "value": 71}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": 0.7552214296770471, "value": 1530}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "High", "coverage": 0.85, "rank": 13, "score": 62}, "method": "attention-ranking-v13", "rank": 13, "score": 62}}, "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-08-12", "source": {"authors": ["Ankita Rajaram Naik", "Anupama Murthi", "Benjamin Elder", "Siyu Huo", "Raavi Gupta", "Abhinav Jain", "Praveen Venkateswaran", "Abdulhamid Adebayo", "Danish Contractor"], "categories": ["cs.AI"], "id": "2608.12282", "title": "VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies", "type": "arxiv", "url": "https://arxiv.org/abs/2608.12282"}, "sourceUpdatedAt": "2026-08-12T17:27:27Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Enterprise agent evaluations need compositional multi-source reasoning with policy constraints; VAKRA offers a public leaderboard and reproducible harness."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
