{"url":"/sota/question-answering-on-piqa","task":{"name":"Question Answering","url":"/task/question-answering","note":null},"dataset":{"name":"PIQA","url":"/dataset/piqa"},"category":"Natural Language Processing","categories":["Miscellaneous","Natural Language Processing","Reasoning"],"category_note":null,"description":"Question answering can be segmented into domain-specific tasks like community question answering and knowledge-base question answering. Popular benchmark datasets for evaluation question answering systems include [SQuAD](/dataset/squad), [HotPotQA](/dataset/hotpotqa), [bAbI](/dataset/babi-1), [TriviaQA](/dataset/triviaqa), [WikiQA](/dataset/wikiqa), and many others. Models for question answering are typically evaluated on metrics like EM and F1. Some recent top performing models are T5 and XLNet.\r\n\r\n( Image credit: [SQuAD](https://rajpurkar.github.io/mlx/qa-and-squad/) )","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Accuracy"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy":"higher"}},"counts":{"rows":67,"rows_with_code":62,"rows_with_paper_page":67,"rows_dated":67,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Unicorn 11B (fine-tuned)","metrics":{"Accuracy":"90.1"},"uses_additional_data":false,"paper_date":"2021-03-24","paper":"/paper/unicorn-on-rainbow-a-universal-commonsense","paper_url":"https://arxiv.org/abs/2103.13009v1","paper_title":"UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark","code":"https://github.com/allenai/rainbow","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":2,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"LLaMA3 8B+MoSLoRA","metrics":{"Accuracy":"89.7"},"uses_additional_data":false,"paper_date":"2024-06-16","paper":"/paper/mixture-of-subspaces-in-low-rank-adaptation","paper_url":"https://arxiv.org/abs/2406.11909v3","paper_title":"Mixture-of-Subspaces in Low-Rank Adaptation","code":"https://github.com/wutaiqiang/moslora","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":2,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":3,"model":"CompassMTL 567M with Tailor","metrics":{"Accuracy":"88.3"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"LLaMA-3 8B + MixLoRA","metrics":{"Accuracy":"87.6"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"DeBERTa-Large 304M","metrics":{"Accuracy":"87.4"},"uses_additional_data":false,"paper_date":"2022-10-29","paper":"/paper/two-is-better-than-many-binary-classification","paper_url":"https://arxiv.org/abs/2210.16495v1","paper_title":"Two is Better than Many? Binary Classification as an Effective Approach to Multi-Choice Question Answering","code":"https://github.com/declare-lab/team","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"CompassMTL 567M","metrics":{"Accuracy":"87.3"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"LLaMA-2 13B + MixLoRA","metrics":{"Accuracy":"86.8"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"Shakti-LLM (2.5B)","metrics":{"Accuracy":"86.2"},"uses_additional_data":false,"paper_date":"2024-10-15","paper":"/paper/shakti-a-2-5-billion-parameter-small-language","paper_url":"https://arxiv.org/abs/2410.11331v1","paper_title":"SHAKTI: A 2.5 Billion Parameter Small Language Model Optimized for Edge AI and Low-Resource Environments","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"DeBERTa-Large 304M (classification-based)","metrics":{"Accuracy":"85.9"},"uses_additional_data":false,"paper_date":"2022-10-29","paper":"/paper/two-is-better-than-many-binary-classification","paper_url":"https://arxiv.org/abs/2210.16495v1","paper_title":"Two is Better than Many? Binary Classification as an Effective Approach to Multi-Choice Question Answering","code":"https://github.com/declare-lab/team","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"ExDeBERTa 567M","metrics":{"Accuracy":"85.5"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"UnifiedQA 3B","metrics":{"Accuracy":"85.3"},"uses_additional_data":false,"paper_date":"2020-05-02","paper":"/paper/unifiedqa-crossing-format-boundaries-with-a","paper_url":"https://arxiv.org/abs/2005.00700v3","paper_title":"UnifiedQA: Crossing Format Boundaries With a Single QA System","code":"https://github.com/allenai/unifiedqa","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":3,"n_samples":7,"n_pointer_only_licence":1}},{"rank_in_archive_order":12,"model":"PaLM 2-L (1-shot)","metrics":{"Accuracy":"85.0"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":13,"model":"Mixtral 8x7B (0-shot)","metrics":{"Accuracy":"83.6"},"uses_additional_data":false,"paper_date":"2024-01-08","paper":"/paper/mixtral-of-experts","paper_url":"https://arxiv.org/abs/2401.04088v1","paper_title":"Mixtral of Experts","code":"https://github.com/jingyaogong/minimind","n_code_links":6,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"PaLM 2-M (1-shot)","metrics":{"Accuracy":"83.2"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"LLaMA-2 7B + MixLoRA","metrics":{"Accuracy":"83.2"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"Mistral 7B (0-shot)","metrics":{"Accuracy":"83.0"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/mistral-7b","paper_url":"https://arxiv.org/abs/2310.06825v1","paper_title":"Mistral 7B","code":"https://github.com/mistralai/mistral-src","n_code_links":6,"syntology":{"n_ran":10,"n_unverified":1,"n_samples":11,"n_pointer_only_licence":1}},{"rank_in_archive_order":17,"model":"LLaMA 65B (0-shot)","metrics":{"Accuracy":"82.8"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":18,"model":"LLaMA 2 70B (0-shot)","metrics":{"Accuracy":"82.8"},"uses_additional_data":false,"paper_date":"2023-07-18","paper":"/paper/llama-2-open-foundation-and-fine-tuned-chat","paper_url":"https://arxiv.org/abs/2307.09288v2","paper_title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","code":"https://github.com/facebookresearch/llama","n_code_links":19,"syntology":{"n_ran":31,"n_unverified":21,"n_samples":52,"n_pointer_only_licence":16}},{"rank_in_archive_order":19,"model":"Camelidae-8×34B","metrics":{"Accuracy":"82.7"},"uses_additional_data":false,"paper_date":"2024-01-05","paper":"/paper/parameter-efficient-sparsity-crafting-from","paper_url":"https://arxiv.org/abs/2401.02731v4","paper_title":"Parameter-Efficient Sparsity Crafting from Dense to Mixture-of-Experts for Instruction Tuning on General Tasks","code":"https://github.com/wuhy68/parameter-efficient-moe","n_code_links":2,"syntology":null},{"rank_in_archive_order":20,"model":"LLaMA 33B (0-shot)","metrics":{"Accuracy":"82.3"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":21,"model":"PaLM 2-S (1-shot)","metrics":{"Accuracy":"82.2"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"Mistral 7B (0-shot)","metrics":{"Accuracy":"82.2"},"uses_additional_data":false,"paper_date":"2024-01-08","paper":"/paper/mixtral-of-experts","paper_url":"https://arxiv.org/abs/2401.04088v1","paper_title":"Mixtral of Experts","code":"https://github.com/jingyaogong/minimind","n_code_links":6,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"MT-NLG 530B (0-shot)","metrics":{"Accuracy":"82.0"},"uses_additional_data":false,"paper_date":"2019-09-17","paper":"/paper/megatron-lm-training-multi-billion-parameter","paper_url":"https://arxiv.org/abs/1909.08053v4","paper_title":"Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism","code":"https://github.com/NVIDIA/Megatron-LM","n_code_links":10,"syntology":{"n_ran":12,"n_unverified":35,"n_samples":47,"n_pointer_only_licence":15}},{"rank_in_archive_order":24,"model":"LLaMA 2 34B (0-shot)","metrics":{"Accuracy":"81.9"},"uses_additional_data":false,"paper_date":"2023-07-18","paper":"/paper/llama-2-open-foundation-and-fine-tuned-chat","paper_url":"https://arxiv.org/abs/2307.09288v2","paper_title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","code":"https://github.com/facebookresearch/llama","n_code_links":19,"syntology":{"n_ran":31,"n_unverified":21,"n_samples":52,"n_pointer_only_licence":16}},{"rank_in_archive_order":25,"model":"Gopher 280B (0-shot)","metrics":{"Accuracy":"81.8"},"uses_additional_data":false,"paper_date":"2021-12-08","paper":"/paper/scaling-language-models-methods-analysis-1","paper_url":"https://arxiv.org/abs/2112.11446v2","paper_title":"Scaling Language Models: Methods, Analysis & Insights from Training Gopher","code":"https://github.com/allenai/dolma","n_code_links":3,"syntology":null},{"rank_in_archive_order":26,"model":"Chinchilla 70B (0-shot)","metrics":{"Accuracy":"81.8"},"uses_additional_data":false,"paper_date":"2022-03-29","paper":"/paper/training-compute-optimal-large-language","paper_url":"https://arxiv.org/abs/2203.15556v1","paper_title":"Training Compute-Optimal Large Language Models","code":"https://github.com/karpathy/llama2.c","n_code_links":2,"syntology":{"n_ran":8,"n_unverified":3,"n_samples":11,"n_pointer_only_licence":4}},{"rank_in_archive_order":27,"model":"FLAN 137B (few-shot, k=10)","metrics":{"Accuracy":"81.7"},"uses_additional_data":false,"paper_date":"2021-09-03","paper":"/paper/finetuned-language-models-are-zero-shot","paper_url":"https://arxiv.org/abs/2109.01652v5","paper_title":"Finetuned Language Models Are Zero-Shot Learners","code":"https://github.com/hiyouga/llama-efficient-tuning","n_code_links":8,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":28,"model":"OPT-175B","metrics":{"Accuracy":"81.07"},"uses_additional_data":false,"paper_date":"2023-01-02","paper":"/paper/massive-language-models-can-be-accurately","paper_url":"https://arxiv.org/abs/2301.00774v3","paper_title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","code":"https://github.com/nvidia/tensorrt-model-optimizer","n_code_links":6,"syntology":{"n_ran":9,"n_unverified":3,"n_samples":12,"n_pointer_only_licence":9}},{"rank_in_archive_order":29,"model":"GPT-3 175B (0-shot)","metrics":{"Accuracy":"81.0"},"uses_additional_data":false,"paper_date":"2020-05-28","paper":"/paper/language-models-are-few-shot-learners","paper_url":"https://arxiv.org/abs/2005.14165v4","paper_title":"Language Models are Few-Shot Learners","code":"https://github.com/ggml-org/llama.cpp","n_code_links":67,"syntology":{"n_ran":41,"n_unverified":24,"n_samples":65,"n_pointer_only_licence":4}},{"rank_in_archive_order":30,"model":"SparseGPT 175B (50% Sparsity)","metrics":{"Accuracy":"80.63"},"uses_additional_data":false,"paper_date":"2023-01-02","paper":"/paper/massive-language-models-can-be-accurately","paper_url":"https://arxiv.org/abs/2301.00774v3","paper_title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","code":"https://github.com/nvidia/tensorrt-model-optimizer","n_code_links":6,"syntology":{"n_ran":9,"n_unverified":3,"n_samples":12,"n_pointer_only_licence":9}},{"rank_in_archive_order":31,"model":"FLAN 137B (0-shot)","metrics":{"Accuracy":"80.5"},"uses_additional_data":false,"paper_date":"2021-09-03","paper":"/paper/finetuned-language-models-are-zero-shot","paper_url":"https://arxiv.org/abs/2109.01652v5","paper_title":"Finetuned Language Models Are Zero-Shot Learners","code":"https://github.com/hiyouga/llama-efficient-tuning","n_code_links":8,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":32,"model":"LLaMA 2 13B (0-shot)","metrics":{"Accuracy":"80.5"},"uses_additional_data":false,"paper_date":"2023-07-18","paper":"/paper/llama-2-open-foundation-and-fine-tuned-chat","paper_url":"https://arxiv.org/abs/2307.09288v2","paper_title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","code":"https://github.com/facebookresearch/llama","n_code_links":19,"syntology":{"n_ran":31,"n_unverified":21,"n_samples":52,"n_pointer_only_licence":16}},{"rank_in_archive_order":33,"model":"LLaMA 13B (0-shot)","metrics":{"Accuracy":"80.1"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":34,"model":"LLaMA 7B (0-shot)","metrics":{"Accuracy":"79.8"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":35,"model":"SparseGPT 175B (4:8 Sparsity)","metrics":{"Accuracy":"79.54"},"uses_additional_data":false,"paper_date":"2023-01-02","paper":"/paper/massive-language-models-can-be-accurately","paper_url":"https://arxiv.org/abs/2301.00774v3","paper_title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","code":"https://github.com/nvidia/tensorrt-model-optimizer","n_code_links":6,"syntology":{"n_ran":9,"n_unverified":3,"n_samples":12,"n_pointer_only_licence":9}},{"rank_in_archive_order":36,"model":"SparseGPT 175B (2:4 Sparsity)","metrics":{"Accuracy":"79.54"},"uses_additional_data":false,"paper_date":"2023-01-02","paper":"/paper/massive-language-models-can-be-accurately","paper_url":"https://arxiv.org/abs/2301.00774v3","paper_title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","code":"https://github.com/nvidia/tensorrt-model-optimizer","n_code_links":6,"syntology":{"n_ran":9,"n_unverified":3,"n_samples":12,"n_pointer_only_licence":9}},{"rank_in_archive_order":37,"model":"RoBERTa-Large 355M","metrics":{"Accuracy":"79.4"},"uses_additional_data":false,"paper_date":"2019-07-26","paper":"/paper/roberta-a-robustly-optimized-bert-pretraining","paper_url":"https://arxiv.org/abs/1907.11692v1","paper_title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","code":"https://github.com/huggingface/transformers","n_code_links":67,"syntology":{"n_ran":37,"n_unverified":11,"n_samples":48,"n_pointer_only_licence":23}},{"rank_in_archive_order":38,"model":"LLaMA 2 7B (0-shot)","metrics":{"Accuracy":"78.8"},"uses_additional_data":false,"paper_date":"2023-07-18","paper":"/paper/llama-2-open-foundation-and-fine-tuned-chat","paper_url":"https://arxiv.org/abs/2307.09288v2","paper_title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","code":"https://github.com/facebookresearch/llama","n_code_links":19,"syntology":{"n_ran":31,"n_unverified":21,"n_samples":52,"n_pointer_only_licence":16}},{"rank_in_archive_order":39,"model":"Bloomberg GPT 50B (1-shot)","metrics":{"Accuracy":"77.9"},"uses_additional_data":false,"paper_date":"2023-03-30","paper":"/paper/bloomberggpt-a-large-language-model-for","paper_url":"https://arxiv.org/abs/2303.17564v3","paper_title":"BloombergGPT: A Large Language Model for Finance","code":"https://github.com/yangletliu/finlora","n_code_links":2,"syntology":null},{"rank_in_archive_order":40,"model":"OPT 66B (1-shot)","metrics":{"Accuracy":"77.6"},"uses_additional_data":false,"paper_date":"2023-03-30","paper":"/paper/bloomberggpt-a-large-language-model-for","paper_url":"https://arxiv.org/abs/2303.17564v3","paper_title":"BloombergGPT: A Large Language Model for Finance","code":"https://github.com/yangletliu/finlora","n_code_links":2,"syntology":null},{"rank_in_archive_order":41,"model":"RoBERTa-large 355M (fine-tuned)","metrics":{"Accuracy":"77.1"},"uses_additional_data":false,"paper_date":"2019-11-26","paper":"/paper/piqa-reasoning-about-physical-commonsense-in","paper_url":"https://arxiv.org/abs/1911.11641v1","paper_title":"PIQA: Reasoning about Physical Commonsense in Natural Language","code":"https://github.com/vered1986/self_talk","n_code_links":2,"syntology":null},{"rank_in_archive_order":42,"model":"phi-1.5-web (1.3B)","metrics":{"Accuracy":"77"},"uses_additional_data":false,"paper_date":"2023-09-11","paper":"/paper/textbooks-are-all-you-need-ii-phi-1-5","paper_url":"https://arxiv.org/abs/2309.05463v1","paper_title":"Textbooks Are All You Need II: phi-1.5 technical report","code":"https://github.com/knowlab/bi-weekly-paper-presentation","n_code_links":1,"syntology":null},{"rank_in_archive_order":43,"model":"BLOOM 176B (1-shot)","metrics":{"Accuracy":"77"},"uses_additional_data":false,"paper_date":"2023-03-30","paper":"/paper/bloomberggpt-a-large-language-model-for","paper_url":"https://arxiv.org/abs/2303.17564v3","paper_title":"BloombergGPT: A Large Language Model for Finance","code":"https://github.com/yangletliu/finlora","n_code_links":2,"syntology":null},{"rank_in_archive_order":44,"model":"Pythia 12B (5-shot)","metrics":{"Accuracy":"76.7"},"uses_additional_data":false,"paper_date":"2023-04-03","paper":"/paper/pythia-a-suite-for-analyzing-large-language","paper_url":"https://arxiv.org/abs/2304.01373v2","paper_title":"Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling","code":"https://github.com/Lightning-AI/lit-gpt","n_code_links":4,"syntology":null},{"rank_in_archive_order":45,"model":"Open-LLaMA-3B-v2","metrics":{"Accuracy":"76.2"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/sheared-llama-accelerating-language-model-pre","paper_url":"https://arxiv.org/abs/2310.06694v2","paper_title":"Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning","code":"https://github.com/princeton-nlp/llm-shearing","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":46,"model":"Pythia 12B (0-shot)","metrics":{"Accuracy":"76"},"uses_additional_data":false,"paper_date":"2023-04-03","paper":"/paper/pythia-a-suite-for-analyzing-large-language","paper_url":"https://arxiv.org/abs/2304.01373v2","paper_title":"Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling","code":"https://github.com/Lightning-AI/lit-gpt","n_code_links":4,"syntology":null},{"rank_in_archive_order":47,"model":"Sheared-LLaMA-2.7B","metrics":{"Accuracy":"75.8"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/sheared-llama-accelerating-language-model-pre","paper_url":"https://arxiv.org/abs/2310.06694v2","paper_title":"Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning","code":"https://github.com/princeton-nlp/llm-shearing","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":48,"model":"GPT-NeoX 20B (1-shot)","metrics":{"Accuracy":"75.8"},"uses_additional_data":false,"paper_date":"2023-03-30","paper":"/paper/bloomberggpt-a-large-language-model-for","paper_url":"https://arxiv.org/abs/2303.17564v3","paper_title":"BloombergGPT: A Large Language Model for Finance","code":"https://github.com/yangletliu/finlora","n_code_links":2,"syntology":null},{"rank_in_archive_order":49,"model":"Pythia 6.9B (0-shot)","metrics":{"Accuracy":"75.2"},"uses_additional_data":false,"paper_date":"2023-04-03","paper":"/paper/pythia-a-suite-for-analyzing-large-language","paper_url":"https://arxiv.org/abs/2304.01373v2","paper_title":"Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling","code":"https://github.com/Lightning-AI/lit-gpt","n_code_links":4,"syntology":null},{"rank_in_archive_order":50,"model":"Sheared-LLaMA-1.3B","metrics":{"Accuracy":"73.4"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/sheared-llama-accelerating-language-model-pre","paper_url":"https://arxiv.org/abs/2310.06694v2","paper_title":"Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning","code":"https://github.com/princeton-nlp/llm-shearing","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":51,"model":"sMLP - deterministic 9.4B (0-shot)","metrics":{"Accuracy":"73"},"uses_additional_data":false,"paper_date":"2022-03-14","paper":"/paper/efficient-language-modeling-with-sparse-all","paper_url":"https://arxiv.org/abs/2203.06850v3","paper_title":"Efficient Language Modeling with Sparse all-MLP","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":52,"model":"GPT-3 Large 760M (0-shot)","metrics":{"Accuracy":"72.9"},"uses_additional_data":false,"paper_date":"2020-05-28","paper":"/paper/language-models-are-few-shot-learners","paper_url":"https://arxiv.org/abs/2005.14165v4","paper_title":"Language Models are Few-Shot Learners","code":"https://github.com/ggml-org/llama.cpp","n_code_links":67,"syntology":{"n_ran":41,"n_unverified":24,"n_samples":65,"n_pointer_only_licence":4}},{"rank_in_archive_order":53,"model":"FLAN-T5-Large 783M","metrics":{"Accuracy":"72.2"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":54,"model":"LaMini-GPT 1.5B","metrics":{"Accuracy":"71.3"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":55,"model":"LaMini-F-T5 783M","metrics":{"Accuracy":"70.6"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":56,"model":"GPT-2-XL 1.5B","metrics":{"Accuracy":"70.5"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":57,"model":"Pythia 1B (5-shot)","metrics":{"Accuracy":"70.4"},"uses_additional_data":false,"paper_date":"2023-04-03","paper":"/paper/pythia-a-suite-for-analyzing-large-language","paper_url":"https://arxiv.org/abs/2304.01373v2","paper_title":"Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling","code":"https://github.com/Lightning-AI/lit-gpt","n_code_links":4,"syntology":null},{"rank_in_archive_order":58,"model":"GPT-2-small 124M (fine-tuned)","metrics":{"Accuracy":"69.2"},"uses_additional_data":false,"paper_date":"2019-11-26","paper":"/paper/piqa-reasoning-about-physical-commonsense-in","paper_url":"https://arxiv.org/abs/1911.11641v1","paper_title":"PIQA: Reasoning about Physical Commonsense in Natural Language","code":"https://github.com/vered1986/self_talk","n_code_links":2,"syntology":null},{"rank_in_archive_order":59,"model":"Gshard 9B","metrics":{"Accuracy":"68.1"},"uses_additional_data":false,"paper_date":"2022-03-14","paper":"/paper/efficient-language-modeling-with-sparse-all","paper_url":"https://arxiv.org/abs/2203.06850v3","paper_title":"Efficient Language Modeling with Sparse all-MLP","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":60,"model":"LaMini-T5 738M","metrics":{"Accuracy":"67.2"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":61,"model":"BERT-large 340M (fine-tuned)","metrics":{"Accuracy":"66.8"},"uses_additional_data":false,"paper_date":"2019-11-26","paper":"/paper/piqa-reasoning-about-physical-commonsense-in","paper_url":"https://arxiv.org/abs/1911.11641v1","paper_title":"PIQA: Reasoning about Physical Commonsense in Natural Language","code":"https://github.com/vered1986/self_talk","n_code_links":2,"syntology":null},{"rank_in_archive_order":62,"model":"BERT-Large 340M","metrics":{"Accuracy":"66.7"},"uses_additional_data":false,"paper_date":"2018-10-11","paper":"/paper/bert-pre-training-of-deep-bidirectional","paper_url":"https://arxiv.org/abs/1810.04805v2","paper_title":"BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding","code":"https://github.com/huggingface/transformers","n_code_links":534,"syntology":{"n_ran":208,"n_unverified":451,"n_samples":659,"n_pointer_only_licence":149}},{"rank_in_archive_order":63,"model":"Base Layers 10B (0-shot)","metrics":{"Accuracy":"63.8"},"uses_additional_data":false,"paper_date":"2022-03-14","paper":"/paper/efficient-language-modeling-with-sparse-all","paper_url":"https://arxiv.org/abs/2203.06850v3","paper_title":"Efficient Language Modeling with Sparse all-MLP","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":64,"model":"HASH Layers 10B (0-shot)","metrics":{"Accuracy":"63.8"},"uses_additional_data":false,"paper_date":"2022-03-14","paper":"/paper/efficient-language-modeling-with-sparse-all","paper_url":"https://arxiv.org/abs/2203.06850v3","paper_title":"Efficient Language Modeling with Sparse all-MLP","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":65,"model":"T5-Large 738M","metrics":{"Accuracy":"55.9"},"uses_additional_data":false,"paper_date":"2023-04-27","paper":"/paper/lamini-lm-a-diverse-herd-of-distilled-models","paper_url":"https://arxiv.org/abs/2304.14402v3","paper_title":"LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions","code":"https://github.com/mbzuai-nlp/lamini-lm","n_code_links":1,"syntology":null},{"rank_in_archive_order":66,"model":"OPT-175B (50% Sparsity)","metrics":{"Accuracy":"54.73"},"uses_additional_data":false,"paper_date":"2023-01-02","paper":"/paper/massive-language-models-can-be-accurately","paper_url":"https://arxiv.org/abs/2301.00774v3","paper_title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","code":"https://github.com/nvidia/tensorrt-model-optimizer","n_code_links":6,"syntology":{"n_ran":9,"n_unverified":3,"n_samples":12,"n_pointer_only_licence":9}},{"rank_in_archive_order":67,"model":"Random chance baseline","metrics":{"Accuracy":"50"},"uses_additional_data":false,"paper_date":"2019-11-26","paper":"/paper/piqa-reasoning-about-physical-commonsense-in","paper_url":"https://arxiv.org/abs/1911.11641v1","paper_title":"PIQA: Reasoning about Physical Commonsense in Natural Language","code":"https://github.com/vered1986/self_talk","n_code_links":2,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":["Accuracy"],"entries":[{"paper":"/paper/arxiv-2609-18131","paper_title":"Colla-Q: Toward Collaborative Experts in MoE Quantization via Minimax Precision Balancing","arxiv_id":"2609.18131","model":"2.54 Colla-Q","row_label":"2.54 Colla-Q","configuration":null,"configuration_in_model":false,"configuration_not_shown_because":null,"values":{"Accuracy":"80.5"},"date":"2026-09-16","month_from_arxiv_id":null,"where_in_paper":"Table 1, row “2.54 Colla-Q”","table_label":"1","table_label_from":"the caption's own label","table_position_1based":1,"caption_head":"Table 1: Zero-shot task performance (%) of the quantized Mixtral 8 × 7B on…","cells":[{"column":"Accuracy","cell_text":"80.5","column_header":"PIQA","value":80.5,"table_index_0based":0,"row":7,"col":3}],"code":"repository linked, no samples harvested"},{"paper":"/paper/arxiv-2606-26587","paper_title":"SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference","arxiv_id":"2606.26587","model":"Llama-3.1-8B SharQ","row_label":"Llama-3.1-8B SharQ","configuration":null,"configuration_in_model":false,"configuration_not_shown_because":null,"values":{"Accuracy":"80.20"},"date":"2026-06-25","month_from_arxiv_id":null,"where_in_paper":"Table 1, row “Llama-3.1-8B SharQ”","table_label":"1","table_label_from":"the caption's own label","table_position_1based":1,"caption_head":"Table 1: Zero-shot and few-shot evaluation of SharQ across three LLMs. Accuracy…","cells":[{"column":"Accuracy","cell_text":"80.20","column_header":"PIQA","value":80.2,"table_index_0based":0,"row":3,"col":5}],"code":"2 of 3 ran"},{"paper":"/paper/arxiv-2605-29843","paper_title":"HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization","arxiv_id":"2605.29843","model":"4 HARP · Llama 2 7B","row_label":"4 HARP","configuration":"Llama 2 7B","configuration_in_model":true,"configuration_not_shown_because":null,"values":{"Accuracy":"78.3"},"date":"2026-05-28","month_from_arxiv_id":null,"where_in_paper":"Table 2, row “4 HARP”","table_label":"2","table_label_from":"the caption's own label","table_position_1based":2,"caption_head":"Table 2: Zero-shot accuracy ( acc , ↑ ) under QuIP# (RHT) vs. HARP on Llama 2…","cells":[{"column":"Accuracy","cell_text":"78.3","column_header":"Llama 2 7B / PIQA","value":78.3,"table_index_0based":1,"row":8,"col":5}],"code":"8 of 9 ran"},{"paper":"/paper/arxiv-2604-13556","paper_title":"YOCO++: Enhancing YOCO with KV Residual Connections for Efficient LLM Inference","arxiv_id":"2604.13556","model":"YOCO++","row_label":"YOCO++","configuration":null,"configuration_in_model":false,"configuration_not_shown_because":null,"values":{"Accuracy":"71.16"},"date":"2026-04-15","month_from_arxiv_id":null,"where_in_paper":"Table 1, row “YOCO++”","table_label":"1","table_label_from":"the caption's own label","table_position_1based":1,"caption_head":"Table 1 : Zero-shot accuracy on commonsense reasoning tasks.","cells":[{"column":"Accuracy","cell_text":"71.16","column_header":"PIQA","value":71.16,"table_index_0based":0,"row":5,"col":6}],"code":"3 of 5 ran"}]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":35,"rows_with_any_sample_ran":33,"distinct_papers_with_graph_line":17,"distinct_papers_with_any_sample_ran":16,"samples_over_distinct_papers":{"n_ran":420,"n_unverified":588,"n_samples":1008,"n_pointer_only_licence":232,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":726,"n_unverified":766,"n_samples":1492,"n_pointer_only_licence":332,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}