| Universal Evasion Attacks on Summarization Scoring | ✓ Link | 48.18 | 19.84 | 45.35 | | Scrambled code + broken (alter) | 2022-10-25 |
| SummaReranker: A Multi-Task Mixture-of-Experts Re-ranking Framework for Abstractive Summarization | ✓ Link | 47.16 | 22.55 | 43.87 | | PEGASUS + SummaReranker | 2022-03-13 |
| Fourier Transformer: Fast Long Range Modeling by Removing Sequence Redundancy with FFT Operator | ✓ Link | 44.76 | 21.55 | 41.34 | | Fourier Transformer | 2023-05-24 |
| GLM: General Language Model Pretraining with Autoregressive Blank Infilling | ✓ Link | 44.7 | 21.4 | 41.4 | | GLM-XXLarge | 2021-03-18 |
| Hierarchical Learning for Generation with Long Source Sequences | | 44.48 | 21.31 | 41.52 | | HAT-BART | 2021-04-15 |
| Extractive Summarization as Text Matching | ✓ Link | 44.41 | 20.86 | 40.55 | | MatchSum (RoBERTa-base) | 2020-04-19 |
| Hie-BART: Document Summarization with Hierarchical BART | | 44.35 | 21.37 | 41.05 | | Hie-BART | 2021-06-01 |
| Extractive Summarization as Text Matching | ✓ Link | 44.22 | 20.62 | 40.38 | | MatchSum (BERT-base) | 2020-04-19 |
| Text Summarization with Pretrained Encoders | ✓ Link | 43.85 | 20.34 | 39.9 | | BertSumExt | 2019-08-22 |
| Big Bird: Transformers for Longer Sequences | ✓ Link | 43.84 | 21.11 | 40.74 | | BigBird-Pegasus | 2020-07-28 |
| Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer | ✓ Link | 43.52 | 21.55 | 40.69 | | T5-11B | 2019-10-23 |
| Fine-tune BERT for Extractive Summarization | ✓ Link | 43.25 | 20.24 | 39.63 | | BERTSUM+Transformer | 2019-03-25 |
| Unified Language Model Pre-training for Natural Language Understanding and Generation | ✓ Link | 43.08 | 20.43 | 40.34 | | UniLM (Abstractive Summarization) | 2019-05-08 |
| Mixture Content Selection for Diverse Sequence Generation | ✓ Link | 41.72 | 18.74 | 38.79 | | Selector+Pointer Generator | 2019-09-04 |
| Neural Document Summarization by Jointly Learning to Score and Select Sentences | ✓ Link | 41.59 | 19.01 | 37.98 | | NeuSUM | 2018-07-06 |
| Bottom-Up Abstractive Summarization | ✓ Link | 41.22 | 18.68 | 38.34 | 32.75 | Bottom-Up Sum | 2018-08-31 |
| Time-aware Large Kernel Convolutions | ✓ Link | 40.59 | 18.97 | 36.81 | | TaLK Convolutions (Deep) | 2020-02-08 |
| Get To The Point: Summarization with Pointer-Generator Networks | ✓ Link | 40.34 | 17.70 | 36.57 | | Lead-3 | 2017-04-14 |
| Time-aware Large Kernel Convolutions | ✓ Link | 40.03 | 18.45 | 36.13 | | TaLK Convolutions (Standard) | 2020-02-08 |
| A Deep Reinforced Model for Abstractive Summarization | ✓ Link | 39.87 | 15.82 | 36.90 | | ML + RL (Paulus et al., 2017) | 2017-05-11 |
| Pay Less Attention with Lightweight and Dynamic Convolutions | ✓ Link | 39.84 | 16.25 | 36.73 | | DynamicConv | 2019-01-29 |
| Pay Less Attention with Lightweight and Dynamic Convolutions | ✓ Link | 39.52 | 15.97 | 36.51 | | LightConv | 2019-01-29 |
| Synthesizer: Rethinking Self-Attention in Transformer Models | ✓ Link | 38.57 | 16.24 | 35.95 | | Synthesizer (R+V) | 2020-05-02 |
| A Deep Reinforced Model for Abstractive Summarization | ✓ Link | 38.30 | 14.81 | 35.49 | | ML + Intra-Attention (Paulus et al., 2017) | 2017-05-11 |
| Coarse-to-Fine Attention Models for Document Summarization | | 31.1 | 15.4 | 28.8 | 23.6 | C2F + ALTERNATE | 2017-09-01 |
| Language Models are Unsupervised Multitask Learners | ✓ Link | 29.34 | 8.27 | 26.58 | | GPT-2 | 2019-02-14 |