Summarization on CNN/Daily Mail original, non-anonymized (test)
41.69ROUGE-1Best Previous Abstractive
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Best Previous Abstractivetype=Abstractive2019.04 | 41.69 | 19.47 | 39.08 | 21 | |
| NEUSUM2018.07 | 41.59 | 19.01 | 37.98 | — | |
| NeuSumMethod Category=Extractive2019.02 | 41.59 | 19.01 | 37.98 | — | |
| BanditSumMethod Category=Extractive2019.02 | 41.5 | 18.7 | 37.6 | — | |
| rnn-ext + RLSummarization Setting=Extractive2018.05 | 41.47 | 18.72 | 37.76 | 22.35 | |
| rnn-ext+RLMethod Category=Extractive2019.02 | 41.47 | 18.72 | 37.76 | — | |
| EditNetMethod Category=Mixed Extractive-Abstractive2019.02 | 41.42 | 19.03 | 38.36 | — | |
| S_D^Ptype=Abstractive, approach=distractor-based pragmatic2019.04 | 41.39 | 18.3 | 38.78 | 21.7 | |
| Rnes w/o coherenceMethod Category=Extractive2019.02 | 41.25 | 18.87 | 37.75 | — | |
| S_R^Ptype=Abstractive, approach=reconstructor-based pragmatic2019.04 | 41.23 | 18.07 | 38.76 | 20.57 | |
| Bottom-upMethod Category=Abstractive2019.02 | 41.22 | 18.68 | 38.34 | — | |
| NN-SE2018.07 | 41.13 | 18.59 | 37.4 | — | |
| LatentMethod Category=Extractive2019.02 | 41.05 | 18.77 | 37.54 | — | |
| rnn-ext + abs + RL + rerankSummarization Setting=Abstractive2018.05 | 40.88 | 17.8 | 38.54 | 20.38 | |
| S0type=Abstractive2019.04 | 40.88 | 17.8 | 38.54 | 20.38 | |
| S0 x2type=Abstractive, ensemble=2 models2019.04 | 40.76 | 17.88 | 38.46 | 19.88 | |
| Inconsistency lossMethod Category=Abstractive2019.02 | 40.68 | 17.97 | 37.13 | — | |
| RL + pg + cbdecCoverage mechanism=true, Decoder=2-decoder (cbdec), Training=Reinforcement Learning2018.09 | 40.66 | 17.87 | 37.06 | 20.51 | |
| RL+pg+cbdecMethod Category=Abstractive2019.02 | 40.66 | 17.87 | 37.06 | — | |
| ff-extSummarization Setting=Extractive2018.05 | 40.63 | 18.35 | 36.82 | 22.91 | |
| CRSUM2018.07 | 40.52 | 18.08 | 36.81 | — | |
| Saliency+Entail.Method Category=Abstractive2019.02 | 40.43 | 18 | 37.1 | — | |
| lead-3Summarization Setting=Extractive2018.05 | 40.34 | 17.7 | 36.57 | 22.21 | |
| Lead-3type=Extractive2019.04 | 40.34 | 17.7 | 36.57 | 22.21 | |
| LEAD3Anonymized=false2018.07 | 40.24 | 17.7 | 36.45 | — | |
| TEXTRANK2018.07 | 40.2 | 17.56 | 36.44 | — | |
| rnn-extSummarization Setting=Extractive2018.05 | 40.17 | 18.11 | 36.41 | 22.81 | |
| pg + cbdecCoverage mechanism=true, Decoder=2-decoder (cbdec)2018.09 | 40.05 | 17.66 | 36.73 | 19.48 | |
| rnn-ext + abs + RLSummarization Setting=Abstractive2018.05 | 40.04 | 17.61 | 37.59 | 21 | |
| rnn-ext+abs+RLMethod Category=Abstractive2019.02 | 40.04 | 17.61 | 37.59 | — | |
| Narayan et al. (2018) (sentence ranking RL)Summarization Setting=Extractive2018.05 | 40 | 18.2 | 36.6 | — | |
| Lead-3Method Category=Extractive2019.02 | 40 | 17.5 | 36.2 | — | |
| RefreshMethod Category=Extractive2019.02 | 40 | 18.2 | 36.6 | — | |
| RL* (Paulus17)Coverage mechanism=false, Anonymized data=true2018.09 | 39.87 | 15.82 | 36.9 | — | |
| Multi-Task(EG+QG)Method Category=Abstractive2019.02 | 39.81 | 17.64 | 36.54 | — | |
| Fan et al. (2017) (controlled)Summarization Setting=Abstractive2018.05 | 39.75 | 17.29 | 36.54 | — | |
| SummaRuNNerMethod Category=Extractive2019.02 | 39.6 | 16.2 | 35.3 | — | |
| RL + pgCoverage mechanism=true, Training=Reinforcement Learning2018.09 | 39.59 | 17.18 | 36.16 | 19.7 | |
| See et al. (2017)Summarization Setting=Abstractive2018.05 | 39.53 | 17.28 | 36.38 | 18.72 | |
| PGN2018.07 | 39.53 | 17.28 | 36.38 | — | |
| pg (See17)Coverage mechanism=true2018.09 | 39.53 | 17.28 | 36.38 | 18.72 | |
| Pointer-GeneratorMethod Category=Abstractive2019.02 | 39.53 | 17.28 | 36.38 | — | |
| ff-ext + absSummarization Setting=Abstractive2018.05 | 39.3 | 17.02 | 36.93 | 20.05 | |
| pg (baseline)Coverage mechanism=true2018.09 | 39.22 | 17.02 | 35.95 | 18.7 | |
| KIGN+Prediction-guideMethod Category=Abstractive2019.02 | 38.95 | 17.12 | 35.68 | — | |
| Inputstype=Extractive2019.04 | 38.93 | 18.23 | 35.9 | 24.66 | |
| RL + pg + cbdecCoverage mechanism=false, Data anonymization=false, Training Objective=Reinforcement Learning, Decoder Type=2-decoder model (closed-book)2018.09 | 38.58 | 16.57 | 35.03 | 18.86 | |
| rnn-ext + absSummarization Setting=Abstractive2018.05 | 38.38 | 16.12 | 36.04 | 19.39 | |
| pg + cbdecCoverage mechanism=false, Data anonymization=false, Decoder Type=2-decoder model (closed-book)2018.09 | 38.21 | 16.45 | 34.7 | 18.37 | |
| RL + pgCoverage mechanism=false, Data anonymization=false, Training Objective=Reinforcement Learning2018.09 | 37.02 | 15.79 | 34 | 17.55 | |
| pg (baseline)Coverage mechanism=false, Data anonymization=false2018.09 | 36.7 | 15.71 | 33.74 | 16.94 | |
| See et al. (2017) (w/o coverage)Summarization Setting=Abstractive2018.05 | 36.44 | 15.66 | 33.42 | 16.65 | |
| pg (See17)Coverage mechanism=false, Data anonymization=false2018.09 | 36.44 | 15.66 | 33.42 | 16.65 | |
| Nallapati et al.Coverage mechanism=false, Data anonymization=true2018.09 | 35.46 | 13.3 | 32.65 | — |