{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.07862", "kind": "paper", "url": "https://arxiv.org/abs/2608.07862"}, {"id": "gh:debobanerjee/surakshaeval", "kind": "repo", "url": "https://github.com/debobanerjee/SurakshaEval"}], "authors": ["Debopriyo Banerjee", "Kapil Rajesh Kavitha", "Angana Borah", "Xudong Han", "Yuxia Wang", "Parameswari Krishnamurthy", "Utkarsh Agarwal", "Atharva Kulkarni", "Swaran Lata", "Ayush Munot", "Dhruv Sahnan", "Aaryamonvikram Singh", "Preslav Nakov", "Monojit Choudhury"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.07862", "source_url": "https://arxiv.org/abs/2608.07862"}, "categories": ["General AI", "Safety", "Safety & Trustworthiness"], "description": {"en": "SurakshaEval evaluates the safety of LLMs across ten major Indian languages and English, using human-written prompts covering generic and region-specific scenarios, with a structured scoring protocol."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.07862", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.07862", "title": "SurakshaEval"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.07862", "modality": null, "name": "SurakshaEval", "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_surakshaeval_6928de7a", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "ed0056e9245310b73a381586a5c9ff8d12ad02c35b3f367fcf6da620c450acc7", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2608.07862"}, "publisher": null, "released": "2026-08-08", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.07862", "source_url": "https://arxiv.org/abs/2608.07862"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-07862", "source": "claire_radar", "source_benchmark_id": "2608.07862", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/debobanerjee/SurakshaEval", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2608.07862", "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": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "e1c84909b838af12685cfca2c2dbae3628bfdb6f910f490a3a3bc8da9212d43e", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "SurakshaEval evaluates the safety of LLMs across ten major Indian languages and English, using human-written prompts covering generic and region-specific scenarios, with a structured scoring protocol.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "To address this gap, we introduce SurakshaEval, a novel safety benchmark composed of human-written prompts spanning real-world scenarios, explicitly designed for ten major Indian languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil, and Telugu, along with English."}, "familyId": "bmf_9fb058d84496", "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/debobanerjee/SurakshaEval", "url": "https://github.com/debobanerjee/SurakshaEval"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/debobanerjee/SurakshaEval", "url": "https://huggingface.co/papers/2608.07862"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/debobanerjee/SurakshaEval.", "url": "https://arxiv.org/abs/2608.07862"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/debobanerjee/SurakshaEval", "url": "https://arxiv.org/abs/2608.07862"}], "scope": "benchmark_repo", "url": "https://github.com/debobanerjee/SurakshaEval", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_surakshaeval_6928de7a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/debobanerjee/SurakshaEval", "data": null, "hfPaper": "https://huggingface.co/papers/2608.07862", "paper": "https://arxiv.org/abs/2608.07862", "pdf": "https://arxiv.org/pdf/2608.07862", "project": null, "report": "https://arxiv.org/abs/2608.07862"}, "metrics": [], "motivation": "Existing safety evaluation datasets for large language models (LLMs) predominantly focus on English and Western contexts, often overlooking the linguistic diversity and culturally grounded safety risks present in other languages.", "name": "SurakshaEval", "oneLine": "SurakshaEval evaluates the safety of LLMs across ten major Indian languages and English, using human-written prompts covering generic and region-specific scenarios, with a structured scoring protocol.", "paperTitle": "SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.07862", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "SurakshaEval Benchmark Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/debobanerjee/SurakshaEval"}], "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": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.0, "value": 0}}, "confidence": "Medium", "coverage": 0.7, "rank": 579, "score": 15}, "method": "attention-ranking-v13", "rank": 579, "score": 15}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-08", "source": {"authors": ["Debopriyo Banerjee", "Kapil Rajesh Kavitha", "Angana Borah", "Xudong Han", "Yuxia Wang", "Parameswari Krishnamurthy", "Utkarsh Agarwal", "Atharva Kulkarni", "Swaran Lata", "Ayush Munot", "Dhruv Sahnan", "Aaryamonvikram Singh", "Preslav Nakov", "Monojit Choudhury"], "categories": ["cs.CL"], "id": "2608.07862", "title": "SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs", "type": "arxiv", "url": "https://arxiv.org/abs/2608.07862"}, "sourceUpdatedAt": "2026-08-08T02:03:33Z", "topics": ["Safety"], "venueAttempts": [], "whyItMatters": "Existing safety benchmarks are largely English-centric, leaving a gap for multilingual and culturally grounded safety assessment. SurakshaEval offers a public protocol to benchmark LLM safety in Indian languages, aiding deployment in diverse linguistic contexts."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
