paper-with-me

Common Sense Reasoning 벤치마크

Common Sense Reasoning on CommonsenseQA

38개 결과 · ⬇ CSV · JSON

Accuracy

20.9 38.81 56.72 74.63 92.54 2018-11 2026-09 BERT-LARGE — 55.9 (2018-11-02) CAGE-reasoning — 64.7 (2019-06-06) RoBERTa-Large 355M — 72.1 (2019-07-26) BERT_CSlarge — 62.2 (2019-08-19) KagNet — 58.9 (2019-09-04) XLNet+GraphReason — 75.3 (2019-09-09) Albert Lan et al. (2020) (ensemble) — 76.5 (2019-09-26) RoBERTa+HyKAS Ma et al. (2019) — 73.2 (2019-10-30) UnifiedQA 11B (fine-tuned) — 79.1 (2020-05-02) T5-XXL 11B (fine-tuned) — 78.1 (2020-05-02) UnifiedQA 11B (zero-shot) — 76.2 (2020-05-02) UnifiedQA 440M (fine-tuned) — 64.0 (2020-05-02) BART-large 440M (fine-tuned) — 62.5 (2020-05-02) DEKCOR — 83.3 (2020-12-09) MUPPET Roberta Large — 79.2 (2021-01-26) Unicorn 11B (fine-tuned) — 79.3 (2021-03-24) QA-GNN — 76.1 (2021-04-13) DeBERTaV3-large+KEAR — 91.2 (2021-12-06) KEAR — 89.4 (2021-12-06) GPT-3 Direct Finetuned — 73.0 (2021-12-06) Chain of thought ASDiv — 28.6 (2022-01-28) STaR (on GPT-J) — 72.3 (2022-03-28) STaR without Rationalization (on GPT-J) — 68.8 (2022-03-28) GPT-J Direct Finetuned — 60.0 (2022-03-28) Few-shot CoT LaMDA 137B — 55.6 (2022-03-28) Few-shot CoT GPT-J — 36.6 (2022-03-28) Few-shot Direct GPT-J — 20.9 (2022-03-28) UL2 20B (chain-of-thought + self-consistency) — 55.7 (2022-05-10) UL2 20B (chain-of-thought) — 51.4 (2022-05-10) UL2 20B (zero-shot) — 34.2 (2022-05-10) DRAGON — 78.2 (2022-10-17) GrapeQA: PEGA — 73.5 (2023-03-22) OPT 66B (1-shot) — 66.4 (2023-03-30) Bloomberg GPT 50B (1-shot) — 65.5 (2023-03-30) BLOOM 176B (1-shot) — 64.2 (2023-03-30) GPT-NeoX 20B (1-shot) — 60.4 (2023-03-30) PaLM 2 (few‑shot, CoT, SC) — 90.4 (2023-05-17) GPT-4o (HPT) — 92.54 (2024-06-18) BERT-LARGE — 55.9 (2018-11-02) CAGE-reasoning — 64.7 (2019-06-06) RoBERTa-Large 355M — 72.1 (2019-07-26) XLNet+GraphReason — 75.3 (2019-09-09) Albert Lan et al. (2020) (ensemble) — 76.5 (2019-09-26) UnifiedQA 11B (fine-tuned) — 79.1 (2020-05-02) DEKCOR — 83.3 (2020-12-09) DeBERTaV3-large+KEAR — 91.2 (2021-12-06) GPT-4o (HPT) — 92.54 (2024-06-18)
RankModel Accuracy Extra Training Data PaperCodeYear
1 GPT-4o (HPT) 92.54 Hierarchical Prompting Taxonomy: A Universal Evaluation Framework for Large Language Models Aligned with Human Cognitive Principles devichand579/HPT 2024
2 DeBERTaV3-large+KEAR 91.2 ✓ Human Parity on CommonsenseQA: Augmenting Self-Attention with External Attention microsoft/DEKCOR-CommonsenseQA · microsoft/kear 2021
3 PaLM 2 (few‑shot, CoT, SC) 90.4 ✓ PaLM 2 Technical Report eternityyw/tram-benchmark 2023
4 KEAR 89.4 ✓ Human Parity on CommonsenseQA: Augmenting Self-Attention with External Attention microsoft/DEKCOR-CommonsenseQA · microsoft/kear 2021
5 DEKCOR 83.3 ✓ Fusing Context Into Knowledge Graph for Commonsense Question Answering microsoft/kear · microsoft/DEKCOR-CommonsenseQA 2020
6 Unicorn 11B (fine-tuned) 79.3 UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark allenai/rainbow 2021
7 MUPPET Roberta Large 79.2 ✓ Muppet: Massive Multi-task Representations with Pre-Finetuning facebook/muppet-roberta-base · facebook/muppet-roberta-large 2021
8 UnifiedQA 11B (fine-tuned) 79.1 ✓ UnifiedQA: Crossing Format Boundaries With a Single QA System allenai/unifiedqa · facebookresearch/metaicl 2020
9 DRAGON 78.2 Deep Bidirectional Language-Knowledge Graph Pretraining michiyasunaga/dragon · HaochenLiu2000/QAP 2022
10 T5-XXL 11B (fine-tuned) 78.1 UnifiedQA: Crossing Format Boundaries With a Single QA System allenai/unifiedqa · facebookresearch/metaicl 2020
11 Albert Lan et al. (2020) (ensemble) 76.5 ALBERT: A Lite BERT for Self-supervised Learning of Language Representations huggingface/transformers · tensorflow/models · PaddlePaddle/PaddleNLP · +45 2019
12 UnifiedQA 11B (zero-shot) 76.2 UnifiedQA: Crossing Format Boundaries With a Single QA System allenai/unifiedqa · facebookresearch/metaicl 2020
13 QA-GNN 76.1 QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering michiyasunaga/qagnn · rucaibox/safe · tobias-opsahl/fact-or-fiction · +3 2021
14 XLNet+GraphReason 75.3 Graph-Based Reasoning over Heterogeneous External Knowledge for Commonsense Question Answering DecstionBack/AAAI_2020_CommonsenseQA 2019
15 GrapeQA: PEGA 73.5 GrapeQA: GRaph Augmentation and Pruning to Enhance Question-Answering 2023
16 RoBERTa+HyKAS Ma et al. (2019) 73.2 Towards Generalizable Neuro-Symbolic Systems for Commonsense Question Answering 2019
17 GPT-3 Direct Finetuned 73.0 Human Parity on CommonsenseQA: Augmenting Self-Attention with External Attention microsoft/DEKCOR-CommonsenseQA · microsoft/kear 2021
18 STaR (on GPT-J) 72.3 STaR: Bootstrapping Reasoning With Reasoning ezelikman/STaR 2022
19 RoBERTa-Large 355M 72.1 RoBERTa: A Robustly Optimized BERT Pretraining Approach huggingface/transformers · pytorch/fairseq · PaddlePaddle/PaddleNLP · +64 2019
20 STaR without Rationalization (on GPT-J) 68.8 STaR: Bootstrapping Reasoning With Reasoning ezelikman/STaR 2022
1–20 / 38 다음 → 페이지당 10 20 50 100