Codesota · Computer Vision · Optical Character Recognition · cnn-/-daily-mailTasks/Computer Vision/Optical Character Recognition
Optical Character Recognition · benchmark dataset · 2020 · EN

cnn-/-daily-mail.

Dataset from Papers With Code

Saturated benchmark· last significant update Jun 2022

ROUGE-based evaluation is saturated. No significant improvements since 2022. Modern summarization uses LLM-as-judge (G-Eval), human preference evaluations, or factual consistency metrics.

Submit a result ↵
§ 01 · Leaderboard

Best published scores.

101 results indexed across 4 metrics. Shaded row marks current SOTA; ties broken by submission date.

Primary
accuracy · higher is better
All metrics
ppl, rouge-1, rouge-2, rouge-l
ppl
2 rows
#ModelOrgSubmittedPaper / codeppl
01Bottom-Up Sum—Aug 2018Bottom-Up Abstractive Summarization · code32.75
02C2F + ALTERNATE—Sep 2017papers-with-code23.60
rouge-1
33 rows
#ModelOrgSubmittedPaper / coderouge-1
01Scrambled code + broken (alter)—Oct 2022Universal Evasion Attacks on Summarization Scoring · code48.18
02BRIOOpenYale NLPMar 2022BRIO: Bringing Order to Abstractive Summarization47.78
03PEGASUS + SummaReranker—Mar 2022SummaReranker: A Multi-Task Mixture-of-Experts Re-rankin… · code47.16
04GPT-3.5-Turbo + TriSum RationaleOpenAIMar 2024TriSum: Learning Summarization Ability from Large Langua…46.70
05BRIDOOpenBRIDO AuthorsFeb 2025BRIDO: Bringing Democratic Order to Abstractive Summariz…45.81
06TriSum-JOpenTriSum AuthorsMar 2024TriSum: Learning Summarization Ability from Large Langua…45.70
07Fourier Transformer—May 2023Fourier Transformer: Fast Long Range Modeling by Removin… · code44.76
08GLM-XXLarge—Mar 2021GLM: General Language Model Pretraining with Autoregress… · code44.70
09HAT-BART—Apr 2021Hierarchical Learning for Generation with Long Source Se…44.48
10MatchSum (RoBERTa-base)—Apr 2020Extractive Summarization as Text Matching · code44.41
11Hie-BART—Jun 2021papers-with-code44.35
12MatchSum (BERT-base)—Apr 2020Extractive Summarization as Text Matching · code44.22
13BertSumExt—Aug 2019Text Summarization with Pretrained Encoders · code43.85
14BigBird-Pegasus—Jul 2020Big Bird: Transformers for Longer Sequences · code43.84
15T5-11BOpenGoogleOct 2019Exploring the Limits of Transfer Learning with a Unified… · code43.52
16SumHiSOpenSumHiS AuthorsJun 2024SumHiS: Extractive Summarization Exploiting Hidden Struc…43.48
17BERTSUM+Transformer—Mar 2019Fine-tune BERT for Extractive Summarization · code43.25
18UniLM (Abstractive Summarization)—May 2019Unified Language Model Pre-training for Natural Language… · code43.08
19Selector+Pointer Generator—Sep 2019Mixture Content Selection for Diverse Sequence Generatio… · code41.72
20NeuSUM—Jul 2018Neural Document Summarization by Jointly Learning to Sco… · code41.59
21Bottom-Up Sum—Aug 2018Bottom-Up Abstractive Summarization · code41.22
22Llama-2-70B-chatOpenMeta AIJul 2025An Evaluation of Large Language Models on Text Summariza…40.98
23TaLK Convolutions (Deep)—Feb 2020Time-aware Large Kernel Convolutions · code40.59
24Lead-3—Apr 2017Get To The Point: Summarization with Pointer-Generator N… · code40.34
25TaLK Convolutions (Standard)—Feb 2020Time-aware Large Kernel Convolutions · code40.03
26ML + RL (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code39.87
27DynamicConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code39.84
28LightConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code39.52
29Synthesizer (R+V)—May 2020Synthesizer: Rethinking Self-Attention in Transformer Mo… · code38.57
30ML + Intra-Attention (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code38.30
31Mistral-7B-Instruct-v0.1OpenMistral AIJul 2025An Evaluation of Large Language Models on Text Summariza…37.44
32C2F + ALTERNATE—Sep 2017papers-with-code31.10
33GPT-2—Feb 2019papers-with-code · code29.34
rouge-2
33 rows
#ModelOrgSubmittedPaper / coderouge-2
01SumHiSOpenSumHiS AuthorsJun 2024SumHiS: Extractive Summarization Exploiting Hidden Struc…32.52
02BRIOOpenYale NLPMar 2022BRIO: Bringing Order to Abstractive Summarization23.75
03GPT-3.5-Turbo + TriSum RationaleOpenAIMar 2024TriSum: Learning Summarization Ability from Large Langua…23.50
04BRIDOOpenBRIDO AuthorsFeb 2025BRIDO: Bringing Democratic Order to Abstractive Summariz…22.95
05TriSum-JOpenTriSum AuthorsMar 2024TriSum: Learning Summarization Ability from Large Langua…22.70
06PEGASUS + SummaReranker—Mar 2022SummaReranker: A Multi-Task Mixture-of-Experts Re-rankin… · code22.55
07T5-11BOpenGoogleOct 2019Exploring the Limits of Transfer Learning with a Unified… · code21.55
08Fourier Transformer—May 2023Fourier Transformer: Fast Long Range Modeling by Removin… · code21.55
09GLM-XXLarge—Mar 2021GLM: General Language Model Pretraining with Autoregress… · code21.40
10Hie-BART—Jun 2021papers-with-code21.37
11HAT-BART—Apr 2021Hierarchical Learning for Generation with Long Source Se…21.31
12BigBird-Pegasus—Jul 2020Big Bird: Transformers for Longer Sequences · code21.11
13MatchSum (RoBERTa-base)—Apr 2020Extractive Summarization as Text Matching · code20.86
14MatchSum (BERT-base)—Apr 2020Extractive Summarization as Text Matching · code20.62
15UniLM (Abstractive Summarization)—May 2019Unified Language Model Pre-training for Natural Language… · code20.43
16BertSumExt—Aug 2019Text Summarization with Pretrained Encoders · code20.34
17BERTSUM+Transformer—Mar 2019Fine-tune BERT for Extractive Summarization · code20.24
18Scrambled code + broken (alter)—Oct 2022Universal Evasion Attacks on Summarization Scoring · code19.84
19NeuSUM—Jul 2018Neural Document Summarization by Jointly Learning to Sco… · code19.01
20TaLK Convolutions (Deep)—Feb 2020Time-aware Large Kernel Convolutions · code18.97
21Selector+Pointer Generator—Sep 2019Mixture Content Selection for Diverse Sequence Generatio… · code18.74
22Bottom-Up Sum—Aug 2018Bottom-Up Abstractive Summarization · code18.68
23TaLK Convolutions (Standard)—Feb 2020Time-aware Large Kernel Convolutions · code18.45
24Lead-3—Apr 2017Get To The Point: Summarization with Pointer-Generator N… · code17.70
25Llama-2-70B-chatOpenMeta AIJul 2025An Evaluation of Large Language Models on Text Summariza…17.23
26Mistral-7B-Instruct-v0.1OpenMistral AIJul 2025An Evaluation of Large Language Models on Text Summariza…16.42
27DynamicConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code16.25
28Synthesizer (R+V)—May 2020Synthesizer: Rethinking Self-Attention in Transformer Mo… · code16.24
29LightConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code15.97
30ML + RL (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code15.82
31C2F + ALTERNATE—Sep 2017papers-with-code15.40
32ML + Intra-Attention (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code14.81
33GPT-2—Feb 2019papers-with-code · code8.27
rouge-l
33 rows
#ModelOrgSubmittedPaper / coderouge-l
01Scrambled code + broken (alter)—Oct 2022Universal Evasion Attacks on Summarization Scoring · code45.35
02BRIOOpenYale NLPMar 2022BRIO: Bringing Order to Abstractive Summarization44.55
03PEGASUS + SummaReranker—Mar 2022SummaReranker: A Multi-Task Mixture-of-Experts Re-rankin… · code43.87
04BRIDOOpenBRIDO AuthorsFeb 2025BRIDO: Bringing Democratic Order to Abstractive Summariz…42.51
05SumHiSOpenSumHiS AuthorsJun 2024SumHiS: Extractive Summarization Exploiting Hidden Struc…42.44
06TriSum-JOpenTriSum AuthorsMar 2024TriSum: Learning Summarization Ability from Large Langua…41.90
07HAT-BART—Apr 2021Hierarchical Learning for Generation with Long Source Se…41.52
08GLM-XXLarge—Mar 2021GLM: General Language Model Pretraining with Autoregress… · code41.40
09Fourier Transformer—May 2023Fourier Transformer: Fast Long Range Modeling by Removin… · code41.34
10Hie-BART—Jun 2021papers-with-code41.05
11BigBird-Pegasus—Jul 2020Big Bird: Transformers for Longer Sequences · code40.74
12GPT-3.5-Turbo + TriSum RationaleOpenAIMar 2024TriSum: Learning Summarization Ability from Large Langua…40.70
13T5-11BOpenGoogleOct 2019Exploring the Limits of Transfer Learning with a Unified… · code40.69
14MatchSum (RoBERTa-base)—Apr 2020Extractive Summarization as Text Matching · code40.55
15MatchSum (BERT-base)—Apr 2020Extractive Summarization as Text Matching · code40.38
16UniLM (Abstractive Summarization)—May 2019Unified Language Model Pre-training for Natural Language… · code40.34
17BertSumExt—Aug 2019Text Summarization with Pretrained Encoders · code39.90
18BERTSUM+Transformer—Mar 2019Fine-tune BERT for Extractive Summarization · code39.63
19Selector+Pointer Generator—Sep 2019Mixture Content Selection for Diverse Sequence Generatio… · code38.79
20Bottom-Up Sum—Aug 2018Bottom-Up Abstractive Summarization · code38.34
21NeuSUM—Jul 2018Neural Document Summarization by Jointly Learning to Sco… · code37.98
22ML + RL (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code36.90
23TaLK Convolutions (Deep)—Feb 2020Time-aware Large Kernel Convolutions · code36.81
24DynamicConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code36.73
25Lead-3—Apr 2017Get To The Point: Summarization with Pointer-Generator N… · code36.57
26LightConv—Jan 2019Pay Less Attention with Lightweight and Dynamic Convolut… · code36.51
27TaLK Convolutions (Standard)—Feb 2020Time-aware Large Kernel Convolutions · code36.13
28Synthesizer (R+V)—May 2020Synthesizer: Rethinking Self-Attention in Transformer Mo… · code35.95
29ML + Intra-Attention (Paulus et al., 2017)—May 2017A Deep Reinforced Model for Abstractive Summarization · code35.49
30C2F + ALTERNATE—Sep 2017papers-with-code28.80
31Llama-2-70B-chatOpenMeta AIJul 2025An Evaluation of Large Language Models on Text Summariza…27.52
32GPT-2—Feb 2019papers-with-code · code26.58
33Mistral-7B-Instruct-v0.1OpenMistral AIJul 2025An Evaluation of Large Language Models on Text Summariza…24.53
Fig 2 · Rows sorted by score within each metric. Shaded row marks SOTA. Dates reflect model or paper release where available, otherwise the date Codesota accessed the source.
§ 04 · Literature

24 papers
tied to this benchmark.

Every paper below corresponds to at least one row in the leaderboard above. Click through for the arXiv preprint and, when available, the reference implementation.

§ 06 · Contribute

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

Submit a result ↵Read submission guide
What a submission needs
  • 01A public checkpoint or API endpoint
  • 02A reproduction script with frozen commit + seed
  • 03Declared evaluation environment (Python, deps)
  • 04One row per metric declared by this dataset
  • 05A contact so we can follow up on discrepancies