Commonsense Generation on CommonGen (test)
38.79METEORHuman
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Human2022.05 | 38.79 | 37.64 | 52.43 | — | — | — | 54.45 | 46.49 | 99.33 | — | — | — | 36.72 | 52.55 | — | — | — | |
| CONTBackbone=T5-base2022.05 | 32.05 | 15.96 | 28.95 | — | — | — | 43.15 | 31.42 | 96.55 | — | — | — | 20.96 | 42.6 | — | — | — | |
| BARTReported Source=[24]2022.05 | 31.83 | 13.98 | 28 | — | — | — | 41.78 | 29.01 | 97.35 | — | — | — | 22.02 | 39.52 | — | — | — | |
| CALM-baseModel size=base2023.06 | 31.4 | 13.88 | 33 | — | — | — | — | 26.4 | — | — | — | — | — | — | — | — | — | |
| CKT-baseModel size=base2023.06 | 31.4 | 13.65 | 33.1 | — | — | — | — | 26.2 | — | — | — | — | — | — | — | — | — | |
| T5-baseModel size=base2023.06 | 31.2 | 12.99 | 32.4 | — | — | — | — | 24.9 | — | — | — | — | — | — | — | — | — | |
| T5Model Size=large, Reported Source=[24]2022.05 | 31.12 | 15.13 | 28.86 | — | — | — | 42.75 | 31.96 | 92.29 | — | — | — | 21.74 | 43.01 | — | — | — | |
| NEUROLOGIC*Supervision Setting=Supervised, Decoding Strategy=greedy2021.12 | 30.8 | 15.2 | 30.8 | — | — | 2.59 | 43.6 | 28.2 | 97.8 | 2.66 | 2.67 | 2.73 | — | — | — | — | — | |
| NEUROLOGIC*Supervision Setting=Supervised, Decoding Strategy=sample2021.12 | 30.8 | 15.3 | 31 | — | — | 2.58 | 43.4 | 27.9 | 97.7 | 2.64 | 2.64 | 2.74 | — | — | — | — | — | |
| NEUROLOGIC*Supervision Setting=Supervised, Decoding Strategy=beam2021.12 | 30.7 | 15.2 | 31 | — | — | 2.6 | 43.2 | 28.2 | 97.6 | 2.68 | 2.67 | 2.76 | — | — | — | — | — | |
| NEUROLOGIC*Supervision Setting=Unsupervised, Decoding Strategy=greedy2021.12 | 30.7 | 15.6 | 29.6 | — | — | 2.7 | 44.3 | 28.6 | 97.1 | 2.78 | 2.7 | 2.77 | — | — | — | — | — | |
| NEUROLOGICSupervision Setting=Supervised2021.12 | 30.2 | 14.7 | 30.3 | — | — | 2.5 | 42.8 | 26.7 | 97.7 | 2.54 | 2.56 | 2.67 | — | — | — | — | — | |
| NEUROLOGICSupervision Setting=Unsupervised2021.12 | 29.5 | 14.4 | 27.5 | — | — | 2.5 | 41.9 | 24.7 | 96.7 | 2.64 | 2.52 | 2.68 | — | — | — | — | — | |
| CBSSupervision Setting=Supervised2021.12 | 28.5 | 12.9 | 27.1 | — | — | 2.23 | 38.8 | 20.6 | 97.6 | 2.27 | 2.35 | 2.51 | — | — | — | — | — | |
| DBASupervision Setting=Supervised2021.12 | 27.7 | 12.4 | 26.3 | — | — | 2.15 | 38.3 | 18.7 | 97.5 | 2.23 | 2.3 | 2.43 | — | — | — | — | — | |
| GPT-2Reported Source=[24]2022.05 | 26.83 | 12.19 | 23.57 | — | — | — | 39.01 | 23.73 | 79.09 | — | — | — | 16.85 | 33.92 | — | — | — | |
| GBSSupervision Setting=Supervised2021.12 | 26.7 | 11.7 | 26.1 | — | — | 2.01 | 38.2 | 18.4 | 97.4 | 2.06 | 2.17 | 2.29 | — | — | — | — | — | |
| TAPModel=ViLT2025.09 | 25.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 41.93 | |
| LocalModel=ViLT2025.09 | 25.31 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 41.44 | |
| T5Model Size=base, Reported Source=[24]2022.05 | 23.94 | 9.4 | 19.87 | — | — | — | 34.56 | 18.54 | 76.67 | — | — | — | 14.63 | 28.76 | — | — | — | |
| FedDAT + Post-trainModel=FLAVA2025.09 | 16.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 39.79 | |
| TAPModel=FLAVA2025.09 | 15.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 35.61 | |
| TSMHSupervision Setting=Unsupervised2021.12 | 14.5 | 3.6 | 15.4 | — | — | 1.63 | 24.7 | 2.2 | 71.5 | 1.85 | 1.92 | 1.95 | — | — | — | — | — | |
| FedDATModel=FLAVA2025.09 | 13.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 37.12 | |
| LocalModel=FLAVA2025.09 | 13.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.69 | |
| Per-FedAvg + Post-trainModel=FLAVA2025.09 | 12.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 32.73 | |
| FedAvg + Post-trainModel=ViLT2025.09 | 12.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 38.78 | |
| Per-FedAvg + Post-trainModel=ViLT2025.09 | 12.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 38.27 | |
| FedDAT + Post-trainModel=ViLT2025.09 | 11.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 38.57 | |
| FedAvgModel=ViLT2025.09 | 11.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 36.79 | |
| FedDATModel=ViLT2025.09 | 10.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 37.38 | |
| Per-FedAvgModel=FLAVA2025.09 | 10.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 32.15 | |
| FedAvg + Post-trainModel=FLAVA2025.09 | 10.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 32.35 | |
| Per-FedAvgModel=ViLT2025.09 | 10.15 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 35.61 | |
| DisentAFL + Post-trainModel=ViLT2025.09 | 9.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 33.76 | |
| DisentAFLModel=ViLT2025.09 | 9.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 32.93 | |
| FedAvgModel=FLAVA2025.09 | 9.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.29 | |
| DisentAFL + Post-trainModel=FLAVA2025.09 | 8.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 27.8 | |
| DisentAFLModel=FLAVA2025.09 | 7.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 26.67 | |
| A*esqueSupervision regime=unsupervised, Base model=GPT2-large2024.06 | — | 15.6 | 29.6 | — | — | — | 44.3 | 28.6 | — | — | — | — | — | — | — | — | — | |
| Ctrl-GSupervision regime=unsupervised, Base model=GPT2-large2024.06 | — | 16.2 | 31.2 | — | — | — | 44.8 | 31.5 | — | — | — | — | — | — | — | 100 | — | |
| Fine-tunedFine-tune on=Common Gen2023.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 39.2 | — | — | |
| GeLaToSupervision regime=unsupervised, Base model=GPT2-large2024.06 | — | 15.5 | 30.3 | — | — | — | 43.8 | 29 | — | — | — | — | — | — | — | 100 | — | |
| GENICLBackbone=GPT-Neo 2.7B2025.05 | — | — | — | — | — | — | 35.3 | — | — | — | — | — | — | — | — | — | — | |
| GENICLBackbone=Vicuna-13B2025.05 | — | — | — | — | — | — | 39.7 | — | — | — | — | — | — | — | — | — | — | |
| GENICLBackbone=Qwen2.5-3B2025.05 | — | — | — | — | — | — | 37.4 | — | — | — | — | — | — | — | — | — | — | |
| GENICLBackbone=LLaMA-3.2-3B2025.05 | — | — | — | — | — | — | 39.8 | — | — | — | — | — | — | — | — | — | — | |
| KGBART2021.07 | — | 168.3 | 32.7 | — | — | 98.6 | — | — | — | — | — | — | — | — | — | — | — | |
| LlamaModel Variant=Base model2023.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 21.14 | — | — | |
| LLM-RBackbone=GPT-Neo 2.7B2025.05 | — | — | — | — | — | — | 34.5 | — | — | — | — | — | — | — | — | — | — | |
| LLM-RBackbone=Vicuna-13B2025.05 | — | — | — | — | — | — | 39.7 | — | — | — | — | — | — | — | — | — | — | |
| LLM-RBackbone=Qwen2.5-3B2025.05 | — | — | — | — | — | — | 35.6 | — | — | — | — | — | — | — | — | — | — | |
| LLM-RBackbone=LLaMA-3.2-3B2025.05 | — | — | — | — | — | — | 39.1 | — | — | — | — | — | — | — | — | — | — | |
| LM-Cocktail10Fine-tune on=Common Gen2023.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 41.45 | — | — | |
| LM-Cocktail2Fine-tune on=Common Gen2023.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 41.22 | — | — | |
| T5-BaseBackbone=T5-Base2021.07 | — | 159.5 | 31.9 | 95.4 | 93.3 | 94 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-Base + GBackbone=T5-Base, Constraint=Ground Truth (G)2021.07 | — | 110.7 | 27.8 | 100 | 100 | 100 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-Base + NRETM PcBackbone=T5-Base, Constraint=Pc2021.07 | — | 164.4 | 32.1 | 95.7 | 94.6 | 95.6 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-Base + NRETM PeBackbone=T5-Base, Constraint=Pe2021.07 | — | 169.8 | 32.7 | 99.5 | 99.1 | 99.3 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-LargeBackbone=T5-Large2021.07 | — | 163.6 | 32.4 | 94.4 | 93 | 93.9 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-Large + NRETM PcBackbone=T5-Large, Constraint=Pc2021.07 | — | 167.3 | 33 | 93.9 | 93.8 | 93.9 | — | — | — | — | — | — | — | — | — | — | — | |
| T5-Large + NRETM PeBackbone=T5-Large, Constraint=Pe2021.07 | — | 175 | 33.8 | 99.6 | 99.2 | 99.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotBackbone=GPT-Neo 2.7B2025.05 | — | — | — | — | — | — | 14.2 | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotBackbone=Vicuna-13B2025.05 | — | — | — | — | — | — | 17.4 | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotBackbone=Qwen2.5-3B2025.05 | — | — | — | — | — | — | 23.6 | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotBackbone=LLaMA-3.2-3B2025.05 | — | — | — | — | — | — | 16.2 | — | — | — | — | — | — | — | — | — | — |