{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:AY7B3MSLIICKEHJXAFZ7D62B7J","short_pith_number":"pith:AY7B3MSL","schema_version":"1.0","canonical_sha256":"063e1db24b4204a21d370173f1fb41fa4a41edf019b497bd0f153385a2b3ba48","source":{"kind":"arxiv","id":"2306.09212","version":2},"attestation_state":"computed","paper":{"title":"CMMLU: Measuring massive multitask language understanding in Chinese","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Most large language models score below 50 percent on a new Chinese multitask understanding benchmark.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fajri Koto, Hai Zhao, Haonan Li, Nan Duan, Timothy Baldwin, Yeyun Gong, Yifei Yang, Yixuan Zhang","submitted_at":"2023-06-15T15:49:51Z","abstract_excerpt":"As the capabilities of large language models (LLMs) continue to advance, evaluating their performance becomes increasingly crucial and challenging. This paper aims to bridge this gap by introducing CMMLU, a comprehensive Chinese benchmark that covers various subjects, including natural science, social sciences, engineering, and humanities. We conduct a thorough evaluation of 18 advanced multilingual- and Chinese-oriented LLMs, assessing their performance across different subjects and settings. The results reveal that most existing LLMs struggle to achieve an average accuracy of 50%, even when "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":true},"canonical_record":{"source":{"id":"2306.09212","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-15T15:49:51Z","cross_cats_sorted":[],"title_canon_sha256":"85638d27497b05a8b6e6cfc2333eaaca9948e66c407771973cfd620a8c90441e","abstract_canon_sha256":"d9c30447ea56c65bd128e617eaa5ad9b4e951a27c9baad2ab43a562a40a37796"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:46.588295Z","signature_b64":"k66DnZfs4bEnjYNQjeBZetig6fSkrRdVvjGp4NKFXffYv0dA5oS/kBtsQijvM9Lb60FaetuDx+juPyNngLB1DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"063e1db24b4204a21d370173f1fb41fa4a41edf019b497bd0f153385a2b3ba48","last_reissued_at":"2026-05-17T23:38:46.587873Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:46.587873Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CMMLU: Measuring massive multitask language understanding in Chinese","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Most large language models score below 50 percent on a new Chinese multitask understanding benchmark.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fajri Koto, Hai Zhao, Haonan Li, Nan Duan, Timothy Baldwin, Yeyun Gong, Yifei Yang, Yixuan Zhang","submitted_at":"2023-06-15T15:49:51Z","abstract_excerpt":"As the capabilities of large language models (LLMs) continue to advance, evaluating their performance becomes increasingly crucial and challenging. This paper aims to bridge this gap by introducing CMMLU, a comprehensive Chinese benchmark that covers various subjects, including natural science, social sciences, engineering, and humanities. We conduct a thorough evaluation of 18 advanced multilingual- and Chinese-oriented LLMs, assessing their performance across different subjects and settings. The results reveal that most existing LLMs struggle to achieve an average accuracy of 50%, even when "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"most existing LLMs struggle to achieve an average accuracy of 50%, even when provided with in-context examples and chain-of-thought prompts, whereas the random baseline stands at 25%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The questions in CMMLU accurately represent the knowledge and reasoning demands of real Chinese-language tasks across the covered subjects.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CMMLU benchmark shows most advanced LLMs score below 50% accuracy on Chinese multitask understanding, well above the 25% random baseline but revealing major room for improvement.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Most large language models score below 50 percent on a new Chinese multitask understanding benchmark.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"afd9388db84a4af3ea61c6f5c90367cea2a2ecbba52536f98631f829755bf1fe"},"source":{"id":"2306.09212","kind":"arxiv","version":2},"verdict":{"id":"7e67a584-1eaf-49f9-b182-dc341b1f0120","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T20:58:16.649845Z","strongest_claim":"most existing LLMs struggle to achieve an average accuracy of 50%, even when provided with in-context examples and chain-of-thought prompts, whereas the random baseline stands at 25%.","one_line_summary":"CMMLU benchmark shows most advanced LLMs score below 50% accuracy on Chinese multitask understanding, well above the 25% random baseline but revealing major room for improvement.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The questions in CMMLU accurately represent the knowledge and reasoning demands of real Chinese-language tasks across the covered subjects.","pith_extraction_headline":"Most large language models score below 50 percent on a new Chinese multitask understanding benchmark."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"9600d38a99db9286fc9a63dad72fb1478b226bdd6e157f3b3acdfdb5cd52bdde"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2306.09212","created_at":"2026-05-17T23:38:46.587936+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.09212v2","created_at":"2026-05-17T23:38:46.587936+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.09212","created_at":"2026-05-17T23:38:46.587936+00:00"},{"alias_kind":"pith_short_12","alias_value":"AY7B3MSLIICK","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_16","alias_value":"AY7B3MSLIICKEHJX","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_8","alias_value":"AY7B3MSL","created_at":"2026-05-18T12:33:33.725879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":18,"internal_anchor_count":18,"sample":[{"citing_arxiv_id":"2502.11089","citing_title":"Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention","ref_index":68,"is_internal_anchor":true},{"citing_arxiv_id":"2412.18925","citing_title":"HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs","ref_index":86,"is_internal_anchor":true},{"citing_arxiv_id":"2512.15745","citing_title":"LLaDA2.0: Scaling Up Diffusion Language Models to 100B","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2403.04652","citing_title":"Yi: Open Foundation Models by 01.AI","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11663","citing_title":"Human-Grounded Multimodal Benchmark with 900K-Scale Aggregated Student Response Distributions from Japan's National Assessment of Academic Ability","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2601.02780","citing_title":"MiMo-V2-Flash Technical Report","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27393","citing_title":"MiniCPM-o 4.5: Towards Real-Time Full-Duplex Omni-Modal Interaction","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08738","citing_title":"SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2401.06066","citing_title":"DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models","ref_index":33,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24690","citing_title":"Can LLMs Act as Historians? Evaluating Historical Research Capabilities of LLMs via the Chinese Imperial Examination","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2604.18946","citing_title":"Reasoning Structure Matters for Safety Alignment of Reasoning Models","ref_index":56,"is_internal_anchor":true},{"citing_arxiv_id":"2401.02954","citing_title":"DeepSeek LLM: Scaling Open-Source Language Models with Longtermism","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2405.04434","citing_title":"DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2502.09992","citing_title":"Large Language Diffusion Models","ref_index":124,"is_internal_anchor":true},{"citing_arxiv_id":"2507.20534","citing_title":"Kimi K2: Open Agentic Intelligence","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2504.10479","citing_title":"InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models","ref_index":63,"is_internal_anchor":true},{"citing_arxiv_id":"2412.05271","citing_title":"Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling","ref_index":127,"is_internal_anchor":true},{"citing_arxiv_id":"2508.18265","citing_title":"InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency","ref_index":61,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J","json":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J.json","graph_json":"https://pith.science/api/pith-number/AY7B3MSLIICKEHJXAFZ7D62B7J/graph.json","events_json":"https://pith.science/api/pith-number/AY7B3MSLIICKEHJXAFZ7D62B7J/events.json","paper":"https://pith.science/paper/AY7B3MSL"},"agent_actions":{"view_html":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J","download_json":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J.json","view_paper":"https://pith.science/paper/AY7B3MSL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.09212&json=true","fetch_graph":"https://pith.science/api/pith-number/AY7B3MSLIICKEHJXAFZ7D62B7J/graph.json","fetch_events":"https://pith.science/api/pith-number/AY7B3MSLIICKEHJXAFZ7D62B7J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J/action/storage_attestation","attest_author":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J/action/author_attestation","sign_citation":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J/action/citation_signature","submit_replication":"https://pith.science/pith/AY7B3MSLIICKEHJXAFZ7D62B7J/action/replication_record"}},"created_at":"2026-05-17T23:38:46.587936+00:00","updated_at":"2026-05-17T23:38:46.587936+00:00"}