Browse › Computer Vision › Sentiment Analysis › SST-2 Binary classification
SST-2 Binary classification Benchmark (Sentiment Analysis)
Sentiment Analysis is the task of classifying the polarity of a given text. For instance, a text-based tweet can be categorized into either "positive", "negative", or "neutral". Given the text and accompanying labels, a model can be trained to predict the correct sentiment.
Sentiment Analysis techniques can be categorized into machine learning approaches, lexicon-based approaches, and even hybrid methods. Some subcategories of research in sentiment analysis include: multimodal sentiment analysis, aspect-based sentiment analysis, fine-grained opinion analysis, language specific sentiment analysis.
More recently, deep learning techniques, such as RoBERTa and T5, are used to train high-performing sentiment classifiers that are evaluated using metrics like F1, recall, and precision. To evaluate sentiment analysis systems, benchmark datasets like SST, GLUE, and IMDB movie reviews are used.
Further readings:
- Sentiment Analysis Based on Deep Learning: A Comparative Study
The archive carries no text for this table; the description above is the archive's text for the task Sentiment Analysis. archive 2025-07-28
Over time archive 2025-07-28
The chart needs JavaScript; the table below carries every value.
Direction inferred from the metric name, not from the archive: Accuracy (higher is better), Dev Accuracy (higher is better), Attack Success Rate (higher is better). Points are placed at the row's paper date; 87 of 87 rows carry one.
Results archive 2025-07-28
Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.
| Paper | Code | Ran Syntology | Report | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | T5-11B | 97.5 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | ||
| 2 | MT-DNN-SMART | 97.5 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 3 | T5-3B | 97.4 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | ||
| 4 | MUPPET Roberta Large | 97.4 | – | Paper | Code | 2021 | linked, not harvested | report | ||
| 5 | ALBERT | 97.1 | – | Paper | Code | 2019 | 46 of 126 ran · 80 unverified | report | ||
| 6 | StructBERTRoBERTa ensemble | 97.1 | – | Paper | – | 2019 | no code linked | report | ||
| 7 | XLNet (single model) | 97 | – | Paper | Code | 2019 | 10 of 24 ran · 14 unverified | report | ||
| 8 | ELECTRA | 96.9 | – | Paper | Code | 2020 | 26 of 40 ran · 14 unverified | report | ||
| 9 | RoBERTa-large 355M + Entailment as Few-shot Learner | 96.9 | – | Paper | Code | 2021 | 1 of 3 ran · 2 unverified | report | ||
| 10 | XLNet-Large (ensemble) | 96.8 | – | Paper | Code | 2019 | 10 of 24 ran · 14 unverified | report | ||
| 11 | FLOATER-large | 96.7 | – | Paper | Code | 2020 | 3 of 6 ran · 3 unverified | report | ||
| 12 | MUPPET Roberta base | 96.7 | – | Paper | Code | 2021 | linked, not harvested | report | ||
| 13 | RoBERTa (ensemble) | 96.7 | – | Paper | Code | 2019 | 22 of 48 ran · 26 unverified | report | ||
| 14 | DeBERTa (large) | 96.5 | – | Paper | Code | 2020 | 4 of 13 ran · 9 unverified | report | ||
| 15 | MT-DNN-ensemble | 96.5 | – | Paper | Code | 2019 | linked, not harvested | report | ||
| 16 | RoBERTa-large 355M (MLP quantized vector-wise, fine-tuned) | 96.4 | – | Paper | Code | 2022 | 2 of 5 ran · 3 unverified | report | ||
| 17 | ASA + RoBERTa | 96.3 | – | Paper | Code | 2022 | linked, not harvested | report | ||
| 18 | T5-Large 770M | 96.3 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | ||
| 19 | Snorkel MeTaL(ensemble) | 96.2 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 20 | PSQ (Chen et al., 2020) | 96.2 | – | Paper | Code | 2020 | 1 of 4 ran · 3 unverified | report | ||
| 21 | Heinsen Routing + RoBERTa-large | 96.0 | ✓ | Paper | Code | 2022 | linked, not harvested | report | ||
| 22 | MT-DNN | 95.6 | – | Paper | Code | 2019 | 5 of 13 ran · 8 unverified | report | ||
| 23 | Heinsen Routing + GPT-2 | 95.6 | ✓ | Paper | Code | 2019 | linked, not harvested | report | ||
| 24 | T5-Base | 95.2 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | ||
| 25 | ERNIE 2.0 Base | 95 | – | Paper | Code | 2019 | 0 of 1 ran · 1 unverified | report | ||
| 26 | RoBERTa+DualCL | 94.91 | – | Paper | Code | 2022 | linked, not harvested | report | ||
| 27 | BERT-LARGE | 94.9 | – | Paper | Code | 2018 | 204 of 659 ran · 455 unverified | report | ||
| 28 | RoBERTa + SubRegWeigh (K-means) | 94.84 | – | Paper | Code | 2024 | linked, not harvested | report | ||
| 29 | SpanBERT | 94.8 | – | Paper | Code | 2019 | 3 of 15 ran · 12 unverified | report | ||
| 30 | gMLP-large | 94.8 | – | Paper | Code | 2021 | 34 of 44 ran · 10 unverified | report | ||
| 31 | Q-BERT (Shen et al., 2020) | 94.8 | – | Paper | – | 2019 | no code linked | report | ||
| 32 | Q8BERT (Zafrir et al., 2019) | 94.7 | – | Paper | Code | 2019 | 3 of 11 ran · 8 unverified | report | ||
| 33 | CNN Large | 94.6 | – | Paper | – | 2019 | no code linked | report | ||
| 34 | BigBird | 94.6 | – | Paper | Code | 2020 | 10 of 15 ran · 5 unverified | report | ||
| 35 | MLM+ del-word+ reorder | 94.5 | – | Paper | – | 2020 | no code linked | report | ||
| 36 | ASA + BERT-base | 94.1 | – | Paper | Code | 2022 | linked, not harvested | report | ||
| 37 | RealFormer | 94.04 | – | Paper | Code | 2020 | linked, not harvested | report | ||
| 38 | FNet-Large | 94 | – | Paper | Code | 2021 | 2 of 2 ran · 0 unverified | report | ||
| 39 | MT-DNN | 93.6 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 40 | ERNIE | 93.5 | – | Paper | Code | 2019 | 3 of 3 ran · 0 unverified | report | ||
| 41 | Block-sparse LSTM | 93.2 | – | Paper | Code | 2017 | linked, not harvested | report | ||
| 42 | LM-CPPF RoBERTa-base | 93.2 | – | Paper | Code | 2023 | linked, not harvested | report | ||
| 43 | TinyBERT-6 67M | 93.1 | – | Paper | Code | 2019 | 0 of 4 ran · 4 unverified | report | ||
| 44 | 24hBERT | 93.0 | – | Paper | Code | 2021 | linked, not harvested | report | ||
| 45 | SMART+BERT-BASE | 93 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 46 | TinyBERT-4 14.5M | 92.6 | – | Paper | Code | 2019 | 0 of 4 ran · 4 unverified | report | ||
| 47 | bmLSTM | 91.8 | – | Paper | Code | 2017 | 0 of 1 ran · 1 unverified | report | ||
| 48 | T5-Small | 91.8 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | ||
| 49 | byte mLSTM7 | 91.7 | – | Paper | Code | 2018 | 2 of 3 ran · 1 unverified | report | ||
| 50 | PAR BERT Base | 91.6 | – | Paper | Code | 2020 | linked, not harvested | report | ||
| 51 | Charformer-Base | 91.6 | – | Paper | Code | 2021 | 7 of 10 ran · 3 unverified | report | ||
| 52 | SqueezeBERT | 91.4 | – | Paper | Code | 2020 | 0 of 1 ran · 1 unverified | report | ||
| 53 | Nyströmformer | 91.4 | – | Paper | Code | 2021 | 1 of 2 ran · 1 unverified | report | ||
| 54 | Bi-CAS-LSTM | 91.3 | – | Paper | – | 2018 | no code linked | report | ||
| 55 | DistilBERT 66M | 91.3 | – | Paper | Code | 2019 | 19 of 27 ran · 8 unverified | report | ||
| 56 | CNN | 91.2 | – | Paper | Code | 2017 | linked, not harvested | report | ||
| 57 | Suffix BiLSTM | 91.2 | – | Paper | – | 2018 | no code linked | report | ||
| 58 | BERT Base | 91.2 | – | Paper | Code | 2019 | 0 of 2 ran · 2 unverified | report | ||
| 59 | Transformer (finetune) | 90.9 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 60 | Single layer bilstm distilled from BERT | 90.7 | – | Paper | Code | 2019 | 9 of 9 ran · 0 unverified | report | ||
| 61 | BCN+Char+CoVe | 90.3 | – | Paper | Code | 2017 | 3 of 3 ran · 0 unverified | report | ||
| 62 | CNN-RNF-LSTM | 90.0 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 63 | Neural Semantic Encoder | 89.7 | – | Paper | Code | 2016 | linked, not harvested | report | ||
| 64 | BLSTM-2DCNN | 89.5 | – | Paper | Code | 2016 | linked, not harvested | report | ||
| 65 | CNN + Logic rules | 89.3 | – | Paper | Code | 2016 | linked, not harvested | report | ||
| 66 | DMN [ankit16] | 88.6 | – | Paper | Code | 2015 | 2 of 5 ran · 3 unverified | report | ||
| 67 | CNN-multichannel [kim2013] | 88.1 | – | Paper | Code | 2014 | 19 of 77 ran · 58 unverified | report | ||
| 68 | Consistency Tree LSTM with tuned Glove vectors [tai2015improved] | 88.0 | – | Paper | Code | 2015 | 6 of 15 ran · 9 unverified | report | ||
| 69 | C-LSTM | 87.8 | – | Paper | Code | 2015 | linked, not harvested | report | ||
| 70 | MPAD-path | 87.75 | – | Paper | Code | 2019 | linked, not harvested | report | ||
| 71 | Standard DR-AGG | 87.6 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 72 | USE_T+CNN (lrn w.e.) | 87.21 | – | Paper | Code | 2018 | 1 of 22 ran · 21 unverified | report | ||
| 73 | Reverse DR-AGG | 87.2 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 74 | DC-MCNN | 86.99 | – | Paper | – | 2018 | no code linked | report | ||
| 75 | STM+TSED+PT+2L | 86.95 | – | Paper | Code | 2019 | linked, not harvested | report | ||
| 76 | Capsule-B | 86.8 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 77 | 2-layer LSTM [tai2015improved] | 86.3 | – | Paper | Code | 2015 | 6 of 15 ran · 9 unverified | report | ||
| 78 | SWEM-concat | 84.3 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 79 | MV-RNN | 82.9 | – | Paper | Code | 2013 | linked, not harvested | report | ||
| 80 | GloVe+Emo2Vec | 82.3 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 81 | Emo2Vec | 81.2 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 82 | ToWE-CBOW | 78.8 | – | Paper | Code | 2018 | linked, not harvested | report | ||
| 83 | Joined Model Multi-tasking | 54.72 | – | Paper | – | 2017 | no code linked | report | ||
| 84 | SMARTRoBERTa | 96.9 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 85 | SMART-MT-DNN | 96.1 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 86 | SMART-BERT | 93.0 | – | Paper | Code | 2019 | 6 of 8 ran · 2 unverified | report | ||
| 87 | Word+ES (Scratch) | 100 | – | Paper | Code | 2022 | 0 of 19 ran · 19 unverified | report |
All 87 rows shown. 87 link to a paper page on this site; 2 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28
Syntology Ran reads "N of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 47 rows have a graph line, from 35 distinct papers; 40 rows (29 papers) have at least one sample that ran. Counting each paper once: Syntology ran 456 of 1,271 samples; 815 unverified. Separately, 300 of those 1,271 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Read from the graph 2026-09-24. Per-sample status is on the paper page.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections