{"about":{"non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","site":"https://codewithpapers.app","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/dataset/milu","name":"MILU","full_name":"Multi-task Indic Language Understanding Benchmark","description_markdown":"## Overview\r\n\r\nMILU (Multi-task Indic Language Understanding Benchmark) is a comprehensive evaluation dataset designed to assess the performance of Large Language Models (LLMs) across 11 Indic languages. It spans 8 domains and 42 subjects, reflecting both general and culturally specific knowledge from India.\r\n\r\n## Key Features\r\n\r\n- **Languages**: Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu, and English\r\n- **Domains**: 8 diverse domains including Arts & Humanities, Social Sciences, STEM, and more\r\n- **Subjects**: 42 subjects covering a wide range of topics\r\n- **Questions**: ~85,000 multiple-choice questions\r\n- **Cultural Relevance**: Incorporates India-specific knowledge from regional and state-level examinations\r\n\r\n## Dataset Statistics\r\n\r\n| Language | Total Questions | Translated Questions | Avg Words Per Question |\r\n|----------|-----------------|----------------------|------------------------|\r\n| Bengali  | 7138            | 1601                 | 15.72                  |\r\n| Gujarati | 5327            | 2755                 | 16.69                  |\r\n| Hindi    | 15450           | 115                  | 20.63                  |\r\n| Kannada  | 6734            | 1522                 | 12.83                  |\r\n| Malayalam| 4670            | 1534                 | 12.82                  |\r\n| Marathi  | 7424            | 1235                 | 18.8                   |\r\n| Odia     | 5025            | 1452                 | 15.63                  |\r\n| Punjabi  | 4363            | 2341                 | 19.9                   |\r\n| Tamil    | 7059            | 1524                 | 13.32                  |\r\n| Telugu   | 7847            | 1298                 | 16.13                  |\r\n| English  | 14036           | -                    | 22.01                  |\r\n| **Total**| **85073**       | **15377**            | **16.77** (avg)        |\r\n\r\n\r\n\r\n## Dataset Structure\r\n\r\n### Test Set\r\nThe test set consists of the MILU (Multi-task Indic Language Understanding) benchmark, which contains approximately 85,000 multiple-choice questions across 11 Indic languages.\r\n\r\n### Validation Set\r\nThe dataset includes a separate validation set of 9,157 samples that can be used for few-shot examples during evaluation. This validation set was created by sampling from each of the 42 subject tags, which were then condensed into 8 broader domains. This approach ensures a balanced representation across subjects and domains, allowing for consistent few-shot prompting across different models and experiments.\r\n\r\n### Subjects spanning MILU\r\n\r\n\r\n| Domain | Subjects |\r\n|--------|----------|\r\n| Arts & Humanities | Architecture and Design, Arts and Culture, Education, History, Language Studies, Literature and Linguistics, Media and Communication, Music and Performing Arts, Religion and Spirituality |\r\n| Business Studies | Business and Management, Economics, Finance and Investment |\r\n| Engineering & Tech | Energy and Power, Engineering, Information Technology, Materials Science, Technology and Innovation, Transportation and Logistics |\r\n| Environmental Sciences | Agriculture, Earth Sciences, Environmental Science, Geography |\r\n| Health & Medicine | Food Science, Health and Medicine |\r\n| Law & Governance | Defense and Security, Ethics and Human Rights, Law and Ethics, Politics and Governance |\r\n| Math and Sciences | Astronomy and Astrophysics, Biology, Chemistry, Computer Science, Logical Reasoning, Mathematics, Physics |\r\n| Social Sciences | Anthropology, International Relations, Psychology, Public Administration, Social Welfare and Development, Sociology, Sports and Recreation |\r\n\r\n\r\n\r\n## Usage\r\n\r\nSince this is a gated dataset, after your request for accessing the dataset is accepted, you can set your HuggingFace token:\r\n\r\n```bash\r\nexport HF_TOKEN=YOUR_TOKEN_HERE\r\n```\r\n\r\nTo load the MILU dataset for a Language:\r\n\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nlanguage = 'Hindi'\r\n\r\n# Use 'test' split for evaluation & 'validation' split for few-shot\r\nsplit = 'test'\r\n\r\nlanguage_data = load_dataset(\"ai4bharat/MILU\", data_dir=language, split=split, token=True)\r\n\r\nprint(language_data[0])\r\n```\r\n\r\n## Evaluation\r\n\r\nWe evaluated 45 different LLMs on MILU, including:\r\n\r\n- Closed proprietary models (e.g., GPT-4o, Gemini-1.5)\r\n- Open-source multilingual models\r\n- Language-specific fine-tuned models\r\n\r\nKey findings:\r\n\r\n- GPT-4o achieved the highest average accuracy at 72%\r\n- Open multilingual models outperformed language-specific fine-tuned models\r\n- Models performed better in high-resource languages compared to low-resource ones\r\n- Performance was lower in culturally relevant areas (e.g., Arts & Humanities) compared to general fields like STEM\r\n\r\nFor detailed results and analysis, please refer to our paper.\r\n\r\n## Citation\r\n\r\nIf you use MILU in your research, please cite our paper:\r\n\r\n```\r\n@misc{verma2024milumultitaskindiclanguage,\r\n      title={MILU: A Multi-task Indic Language Understanding Benchmark}, \r\n      author={Sshubam Verma and Mohammed Safi Ur Rahman Khan and Vishwajeet Kumar and Rudra Murthy and Jaydeep Sen},\r\n      year={2024},\r\n      eprint={2411.02538},\r\n      archivePrefix={arXiv},\r\n      primaryClass={cs.CL},\r\n      url={https://arxiv.org/abs/2411.02538}, \r\n}```\r\n\r\n## License\r\n\r\nThis dataset is released under the [MIT License](https://opensource.org/licenses/MIT).\r\n\r\n## Contact\r\n\r\nFor any questions or feedback, please contact:\r\n- Sshubam Verma (sshubamverma@ai4bharat.org)\r\n- Mohammed Safi Ur Rahman Khan (safikhan@ai4bharat.org)\r\n- Rudra Murthy (rmurthyv@in.ibm.com)\r\n- Vishwajeet Kumar (vishk024@in.ibm.com)\r\n\r\n## Links\r\n\r\n- [GitHub Repository](https://github.com/AI4Bharat/MILU)\r\n- [Paper](https://arxiv.org/abs/2411.02538)\r\n- [Hugging Face Dataset](https://huggingface.co/datasets/ai4bharat/MILU)","description_withheld":null,"homepage":"https://github.com/AI4Bharat/MILU","introduced_date":"2024-11-04","introduced_date_note":null,"introduced_by":{"paper":"/paper/milu-a-multi-task-indic-language","title":"MILU: A Multi-task Indic Language Understanding Benchmark","first_author":"Sshubam Verma","url":null},"license":{"name":"MIT","url":"https://opensource.org/license/mit"},"modalities":[{"name":"Texts","url":"/datasets/modality/texts"}],"tasks":[{"name":"Question Answering","url":"/task/question-answering","datasets_with_task":"/datasets/task/question-answering"}],"languages":[{"name":"English","url":"/datasets/language/english"},{"name":"Bengali","url":"/datasets/language/bengali"},{"name":"Hindi","url":"/datasets/language/hindi"},{"name":"Marathi","url":"/datasets/language/marathi"},{"name":"Tamil","url":"/datasets/language/tamil"},{"name":"Telugu","url":"/datasets/language/telugu"},{"name":"Gujarati","url":"/datasets/language/gujarati"},{"name":"Kannada","url":"/datasets/language/kannada"},{"name":"Malayalam","url":"/datasets/language/malayalam"},{"name":"Odia","url":"/datasets/language/odia"},{"name":"Punjabi","url":"/datasets/language/punjabi"}],"variants":["MILU"],"data_loaders":[],"num_papers_in_archive":1,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[],"papers":{"order":"date (newest first), then slug","population":"every paper with a leaderboard row on this dataset's benchmarks (the benchmark-backed subset): the archive's own papers-using-this-dataset list was never published, so this is not that list; num_papers_in_archive is the archive's own count","page":1,"pages_in_order":1,"rows_per_page":100,"shown":0,"of":0,"counts":{"papers_with_a_benchmark_row":0,"with_a_code_link":0,"where_syntology_ran_a_sample":0,"not_listed_spam_title":0,"listed":0,"listed_where_code_ran":0,"where_syntology_ran_a_sample_split":{"with_a_run_with_no_instrument_failure":0,"every_run_a_failure_of_syntologys_instrument":0,"listed_with_a_run_with_no_instrument_failure":0,"listed_every_run_a_failure_of_syntologys_instrument":0,"filter":{"states":["a run with no instrument failure","any run, instrument failures included"],"default":"a run with no instrument failure","note":"on the 'only where code ran' pages the default hides, in the browser, the rows where every run was a failure of Syntology's instrument; the second state shows them again. Rows are hidden, never re-ordered; these twins list every row"}},"definition":"distinct papers with at least one leaderboard row on this dataset's benchmarks; 'where Syntology ran a sample' counts papers with at least one harvested sample that ran, which is not a correctness claim"},"items":[],"full_list":"papers_with_a_benchmark_row: every listed paper, in this order"},"papers_with_a_benchmark_row":[],"papers_where_code_ran":0,"papers_where_code_ran_first_page":null,"papers_where_code_ran_caption":"We ran code from the paper's repository; we did not run it on this dataset or check it against this dataset's benchmarks.","pages":[],"row_route":"each paper's syntology object carries sample_list (this site's paper page, #ran), syntology_url (syntology.ai's page for the paper, only when that route probed live) and mcp (the agent call that lists the paper's harvested samples; not a verdict on the paper)","developers":"https://syntology.ai/developers","syntology_totals":{"read_at":"2026-10-05T00:36:11+00:00","papers_with_samples":0,"samples_harvested":0,"samples_ran":0,"samples_constructed":0,"samples_ran_checked":0,"samples_ran_instrument_failed":0,"samples_unverified":0,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":0,"papers_with_a_sample_that_ran":0,"papers_with_a_run_with_no_instrument_failure":0,"papers_where_every_run_was_an_instrument_failure":0,"note":"the per-paper counts of every listed paper (all pages), summed; not a rate"},"key_notes":{"samples_ran_checked":"legacy name, kept unchanged so existing readers do not break: it counts the samples that ran with no instrument failure (honoured, violated, and ran with no contract checked); it does not mean a contract was checked, and the pages print it as 'K with no instrument failure', not 'K checked'","samples_constructed":"a sub-count of the samples that ran, never subtracted from them and never a failure: an executed sample whose run returned an instance of its own class (fixture_out_type equals the entry name): the run built an object and did not compute a result (Syntology's RAN record, counts.constructed)"},"since_archive_judged":{"papers":0,"papers_placed":0,"judged_by":{"model":"a model","models_as_recorded":[],"from":"lb_rows_run_verified.jsonl: each record's `model` field (the extractor's call that wrote decisions[])"},"note":"a model's judgement from the paper's own tables that they report on this leaderboard, not a result: no value from a paper that was not placed is printed or carried here; papers counts each paper once over these leaderboards","backfill":"8,886 of 18,637 newer papers checked so far (733 more had nothing to place; 41 could not be read)","backfill_complete":false,"leaderboards":[]},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks.","record_sha256":"7230df35891248bae8f761763d3fcc6c27a8b405ab5853acb70a924033f29501","record_changed_at":"2026-10-05","record_changed_at_basis":"changed"}