# Sentiment Analysis on SST-2 (test)

Source: [Sentiment Analysis on SST-2 (test)](<https://www.sota2.com/research/sota/sentiment-analysis-on-sst-2-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-04T11:29:29.067Z

Evaluation of Sentiment Analysis performance on the test split of the Stanford Sentiment Treebank 2 (SST-2) dataset, measured by Accuracy.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Sentiment Analysis |
| Dataset (including any stated split) | SST-2 (test) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 162 of 162 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy](<https://www.sota2.com/research/sota/sentiment-analysis-on-sst-2-test.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Previous SOTA | 97.1 | Not recorded | 68fa67921e209131cea1718e |
| RoBERTa + SKEP | 97 | Not recorded | 68fa67921e209131cea1718e |
| RoBERTa + SKEP | 96.7 | Not recorded | 68fa67921e209131cea1718e |
| LM-Cocktail2 | 96.56 | Not recorded | 68fa71661e209131cea5132e |
| LM-Cocktail10 | 96.56 | Not recorded | 68fa71661e209131cea5132e |
| RoBERTa | 96.5 | Not recorded | 68fa67921e209131cea1718e |
| CLS-based FT + TAPT | 96 | Not recorded | 68fa70d11e209131cea4d652 |
| PGI | 95.77 | Not recorded | 68fa6fb71e209131cea46fdb |
| Non-Aligned | 95.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Fine-tuned | 95.53 | Not recorded | 68fa71661e209131cea5132e |
| Prompt-based FT (hard) + PCP | 95.5 | Not recorded | 68fa70d11e209131cea4d652 |
| Prompt-based FT (hard) | 95.2 | Not recorded | 68fa70d11e209131cea4d652 |
| CLS-based FT | 95.1 | Not recorded | 68fa70d11e209131cea4d652 |
| Fine-tuning (full) | 95 | Not recorded | 68fa6e771e209131cea400bc |
| Vaccine | 95 | Not recorded | 68fa75d01e209131cea69fa9 |
| Unpatched | 95 | Not recorded | 6a210685c10484bac23f00f1 |
| RoBERTa | 94.9 | Not recorded | 68fa67921e209131cea1718e |
| SFT | 94.8 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vlguard | 94.8 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vlguard | 94.8 | Not recorded | 68fa75d01e209131cea69fa9 |
| Extra Annotation | 94.7 | Not recorded | 68fa73401e209131cea5b47a |
| Non-Aligned | 94.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Non-Aligned | 94.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vlguard | 94.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vlguard | 94.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vlguard | 94.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| GENICL | 94.6 | Not recorded | 68fa78c21e209131cea7a598 |
| CoTAM | 94.5 | Not recorded | 68fa73401e209131cea5b47a |
| GENICL | 94.5 | Not recorded | 68fa78c21e209131cea7a598 |
| Non-Aligned | 94.4 | Not recorded | 68fa75d01e209131cea69fa9 |
| SFT | 94.4 | Not recorded | 68fa75d01e209131cea69fa9 |
| SFT | 94.4 | Not recorded | 68fa75d01e209131cea69fa9 |
| Prompt-based FT (soft) + PCP | 94.3 | Not recorded | 68fa70d11e209131cea4d652 |
| FlipDA++ | 94.3 | Not recorded | 68fa73401e209131cea5b47a |
| LLM-R | 94.3 | Not recorded | 68fa78c21e209131cea7a598 |
| LLM Pseudo Label | 94.2 | Not recorded | 68fa73401e209131cea5b47a |
| SFT | 94.2 | Not recorded | 68fa75d01e209131cea69fa9 |
| SFT | 94.2 | Not recorded | 68fa75d01e209131cea69fa9 |
| Base | 94 | Not recorded | 68fa73401e209131cea5b47a |
| COTDA | 94 | Not recorded | 68fa73401e209131cea5b47a |
| Non-Aligned | 94 | Not recorded | 68fa75d01e209131cea69fa9 |
| GENICL | 94 | Not recorded | 68fa78c21e209131cea7a598 |
| FT | 94 | Not recorded | 6a210685c10484bac23f00f1 |
| FT\_l2 | 94 | Not recorded | 6a210685c10484bac23f00f1 |
| FT\_linf | 94 | Not recorded | 6a210685c10484bac23f00f1 |
| Patcher | 94 | Not recorded | 6a210685c10484bac23f00f1 |
| Prompt-based FT (soft) + PCP | 93.9 | Not recorded | 68fa70d11e209131cea4d652 |
| KEN | 93.8 | Not recorded | 68fa73851e209131cea5cebc |
| Vaccine | 93.8 | Not recorded | 68fa75d01e209131cea69fa9 |
| LLM-R | 93.8 | Not recorded | 68fa78c21e209131cea7a598 |
| Prompt-based FT (hard) + PCP | 93.6 | Not recorded | 68fa70d11e209131cea4d652 |
| Human Labeled | 93.52 | Not recorded | 68fa6fb71e209131cea46fdb |
| DART | 93.5 | Not recorded | 68fa6e771e209131cea400bc |
| Prompt-based FT (hard) + TAPT | 93.5 | Not recorded | 68fa70d11e209131cea4d652 |
| Bert-base | 93.37 | Not recorded | 68fa73851e209131cea5cebc |
| Hybrid | 93.23 | Not recorded | 68fa73851e209131cea5cebc |
| Vaccine | 93 | Not recorded | 68fa75d01e209131cea69fa9 |
| LLM-R | 93 | Not recorded | 68fa78c21e209131cea7a598 |
| FinePruning | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| BAERASER | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| Mudjacking | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| SPP | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| MEND | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| ROME | 93 | Not recorded | 6a210685c10484bac23f00f1 |
| KEN | 92.9 | Not recorded | 68fa73851e209131cea5cebc |
| GENICL | 92.7 | Not recorded | 68fa78c21e209131cea7a598 |
| Vaccine | 92.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| Vaccine | 92.6 | Not recorded | 68fa75d01e209131cea69fa9 |
| LLM-R | 92.6 | Not recorded | 68fa78c21e209131cea7a598 |
| LKD+Init. | 92.5 | Not recorded | 68fa6f581e209131cea4524c |
| LKD+Init. | 92.5 | Not recorded | 68fa6f581e209131cea4524c |
| LM-BFF | 92.3 | Not recorded | 68fa6e771e209131cea400bc |
| P-Tuning | 92.2 | Not recorded | 68fa6e771e209131cea400bc |
| HybridNT | 92.2 | Not recorded | 68fa73851e209131cea5cebc |
| LKD | 92 | Not recorded | 68fa6f581e209131cea4524c |
| OneShot | 92 | Not recorded | 6a210685c10484bac23f00f1 |
| Hybrid | 91.97 | Not recorded | 68fa73851e209131cea5cebc |
| ERM | 91.85 | Not recorded | 68fa6f741e209131cea45b0c |
| ZoVH | 91.55 | Not recorded | 6a1d11e5dbd7150506f83e0e |
| RETROPROMPT | 91.4 | Not recorded | 68fa709e1e209131cea4c484 |

### Previous SOTA

Accuracy: 97.1. Result ID: 698d3b7e8b517eb4331ff683.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### RoBERTa + SKEP

Accuracy: 97. Result ID: 698d3b7e8b517eb4331ff687.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Implementation](<https://github.com/baidu/Senta>)

Recorded evaluation configuration: {"Model Size":"large"}. This record does not assert independent reproduction.

### RoBERTa + SKEP

Accuracy: 96.7. Result ID: 698d3b7e8b517eb4331ff685.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Implementation](<https://github.com/baidu/Senta>)

Recorded evaluation configuration: {"Model Size":"base"}. This record does not assert independent reproduction.

### LM-Cocktail2

Accuracy: 96.56. Result ID: 6991cde33a6c3148490aafe2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa71661e209131cea5132e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71661e209131cea5132e.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Fine-tune on":"SST2"}. This record does not assert independent reproduction.

### LM-Cocktail10

Accuracy: 96.56. Result ID: 6991cde33a6c3148490aafe3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa71661e209131cea5132e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71661e209131cea5132e.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Fine-tune on":"SST2"}. This record does not assert independent reproduction.

### RoBERTa

Accuracy: 96.5. Result ID: 698d3b7e8b517eb4331ff686.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Model Size":"large"}. This record does not assert independent reproduction.

### CLS-based FT + TAPT

Accuracy: 96. Result ID: 69915b0b3a6c314849079296.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa677a1e209131cea16a90.md>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"CLS-based Fine-tuning","Continued Pre-training":"TAPT"}. This record does not assert independent reproduction.

### PGI

Accuracy: 95.77. Result ID: 69a153d135dfcabb00e64b42.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fb71e209131cea46fdb.md>)

Recorded evaluation configuration: {"Num. of Samples":"1821","Cost (USD)":"7.33","Time (Mins)":"12"}. This record does not assert independent reproduction.

### Non-Aligned

Accuracy: 95.6. Result ID: 6993fc8a3a6c3148491a1478.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"clean","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Fine-tuned

Accuracy: 95.53. Result ID: 6991cde33a6c3148490aafe1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa71661e209131cea5132e.md>)

Recorded evaluation configuration: {"Fine-tune on":"SST2"}. This record does not assert independent reproduction.

### Prompt-based FT (hard) + PCP

Accuracy: 95.5. Result ID: 69915b0b3a6c314849079299.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Implementation](<https://github.com/ZhengxiangShi/PowerfulPromptFT>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"Prompt-based Fine-tuning (hard)","Continued Pre-training":"PCP"}. This record does not assert independent reproduction.

### Prompt-based FT (hard)

Accuracy: 95.2. Result ID: 69915b0b3a6c314849079297.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a101e209131cea25eae.md>)

[Implementation](<https://github.com/princeton-nlp/LM-BFF>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"Prompt-based Fine-tuning (hard)","Continued Pre-training":"None"}. This record does not assert independent reproduction.

### CLS-based FT

Accuracy: 95.1. Result ID: 69915b0b3a6c314849079295.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"CLS-based Fine-tuning","Continued Pre-training":"None"}. This record does not assert independent reproduction.

### Fine-tuning (full)

Accuracy: 95. Result ID: 699031a23a6c314849fecb71.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e771e209131cea400bc.md>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-large","Training Samples (K)":"full","Evaluation Protocol":"Full fine-tuning"}. This record does not assert independent reproduction.

### Vaccine

Accuracy: 95. Result ID: 6993fc8a3a6c3148491a1495.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Implementation](<https://github.com/git-disl/Vaccine>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.2","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Unpatched

Accuracy: 95. Result ID: 6a212e1afbc373592c1c7632.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### RoBERTa

Accuracy: 94.9. Result ID: 698d3b7e8b517eb4331ff684.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67921e209131cea1718e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Model Size":"base"}. This record does not assert independent reproduction.

### SFT

Accuracy: 94.8. Result ID: 6993fc8a3a6c3148491a147f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.05","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vlguard

Accuracy: 94.8. Result ID: 6993fc8a3a6c3148491a1487.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5db89.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"clean","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vlguard

Accuracy: 94.8. Result ID: 6993fc8a3a6c3148491a1488.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5db89.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.01","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Extra Annotation

Accuracy: 94.7. Result ID: 6992b2083a6c31484910cac0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"Increased"}. This record does not assert independent reproduction.

### Non-Aligned

Accuracy: 94.6. Result ID: 6993fc8a3a6c3148491a1479.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.01","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Non-Aligned

Accuracy: 94.6. Result ID: 6993fc8a3a6c3148491a147b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.1","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vlguard

Accuracy: 94.6. Result ID: 6993fc8a3a6c3148491a1489.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5db89.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.05","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vlguard

Accuracy: 94.6. Result ID: 6993fc8a3a6c3148491a148a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5db89.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.1","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vlguard

Accuracy: 94.6. Result ID: 6993fc8a3a6c3148491a148b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5db89.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.2","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### GENICL

Accuracy: 94.6. Result ID: 6995a11a3badfaffe69dcd0b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Implementation](<https://foundation-model-research.github.io/GenICL>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B"}. This record does not assert independent reproduction.

### CoTAM

Accuracy: 94.5. Result ID: 6992b2083a6c31484910cac4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

[Implementation](<https://www.sota2.com/github.com/KomeijiForce/CoTAM>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"3"}. This record does not assert independent reproduction.

### GENICL

Accuracy: 94.5. Result ID: 6995a11a3badfaffe69dcd0e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Implementation](<https://foundation-model-research.github.io/GenICL>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3.2-3B"}. This record does not assert independent reproduction.

### Non-Aligned

Accuracy: 94.4. Result ID: 6993fc8a3a6c3148491a147c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.2","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### SFT

Accuracy: 94.4. Result ID: 6993fc8a3a6c3148491a147e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.01","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### SFT

Accuracy: 94.4. Result ID: 6993fc8a3a6c3148491a1480.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.1","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Prompt-based FT (soft) + PCP

Accuracy: 94.3. Result ID: 69915b0b3a6c31484907929a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Implementation](<https://github.com/ZhengxiangShi/PowerfulPromptFT>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"Prompt-based Fine-tuning (soft)","Continued Pre-training":"PCP"}. This record does not assert independent reproduction.

### FlipDA++

Accuracy: 94.3. Result ID: 6992b2083a6c31484910cac2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319b7.md>)

[Implementation](<https://github.com/zhouj8553/FlipDA>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"3"}. This record does not assert independent reproduction.

### LLM-R

Accuracy: 94.3. Result ID: 6995a11a3badfaffe69dcd0a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71f21e209131cea53b15.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B"}. This record does not assert independent reproduction.

### LLM Pseudo Label

Accuracy: 94.2. Result ID: 6992b2083a6c31484910cac1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"3"}. This record does not assert independent reproduction.

### SFT

Accuracy: 94.2. Result ID: 6993fc8a3a6c3148491a147d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"clean","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### SFT

Accuracy: 94.2. Result ID: 6993fc8a3a6c3148491a1481.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.2","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Base

Accuracy: 94. Result ID: 6992b2083a6c31484910cabf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"3"}. This record does not assert independent reproduction.

### COTDA

Accuracy: 94. Result ID: 6992b2083a6c31484910cac3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73401e209131cea5b47a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"ICL","Number of shots (K)":"3"}. This record does not assert independent reproduction.

### Non-Aligned

Accuracy: 94. Result ID: 6993fc8a3a6c3148491a147a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.05","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### GENICL

Accuracy: 94. Result ID: 6995a11a3badfaffe69dcd08.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Implementation](<https://foundation-model-research.github.io/GenICL>)

Recorded evaluation configuration: {"Backbone":"Vicuna-13B"}. This record does not assert independent reproduction.

### FT

Accuracy: 94. Result ID: 6a212e1afbc373592c1c7633.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### FT\_l2

Accuracy: 94. Result ID: 6a212e1afbc373592c1c7634.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68d31e209131cea1e6ab.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### FT\_linf

Accuracy: 94. Result ID: 6a212e1afbc373592c1c7635.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68d31e209131cea1e6ab.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### Patcher

Accuracy: 94. Result ID: 6a212e1afbc373592c1c763d.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### Prompt-based FT (soft) + PCP

Accuracy: 93.9. Result ID: 69915b0b3a6c31484907929c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Implementation](<https://github.com/ZhengxiangShi/PowerfulPromptFT>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Semi-supervised (16-shot)","Evaluation Protocol":"Prompt-based Fine-tuning (soft)","Continued Pre-training":"PCP"}. This record does not assert independent reproduction.

### KEN

Accuracy: 93.8. Result ID: 6992e54d3a6c314849123a48.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Implementation](<https://github.com/itsmattei/KEN>)

Recorded evaluation configuration: {"Trainable params":"80M"}. This record does not assert independent reproduction.

### Vaccine

Accuracy: 93.8. Result ID: 6993fc8a3a6c3148491a1494.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Implementation](<https://github.com/git-disl/Vaccine>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.1","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### LLM-R

Accuracy: 93.8. Result ID: 6995a11a3badfaffe69dcd07.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71f21e209131cea53b15.md>)

Recorded evaluation configuration: {"Backbone":"Vicuna-13B"}. This record does not assert independent reproduction.

### Prompt-based FT (hard) + PCP

Accuracy: 93.6. Result ID: 69915b0b3a6c31484907929b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Implementation](<https://github.com/ZhengxiangShi/PowerfulPromptFT>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Semi-supervised (16-shot)","Evaluation Protocol":"Prompt-based Fine-tuning (hard)","Continued Pre-training":"PCP"}. This record does not assert independent reproduction.

### Human Labeled

Accuracy: 93.52. Result ID: 69a153d135dfcabb00e64b41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fb71e209131cea46fdb.md>)

Recorded evaluation configuration: {"Num. of Samples":"67349","Cost (USD)":"4800 - 6700","Time (Mins)":"22740"}. This record does not assert independent reproduction.

### DART

Accuracy: 93.5. Result ID: 699031a23a6c314849fecb70.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e771e209131cea400bc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e771e209131cea400bc.md>)

[Implementation](<https://github.com/zjunlp/DART>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-large","Training Samples (K)":16,"Evaluation Protocol":"Few-shot fine-tuning"}. This record does not assert independent reproduction.

### Prompt-based FT (hard) + TAPT

Accuracy: 93.5. Result ID: 69915b0b3a6c314849079298.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70d11e209131cea4d652.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa677a1e209131cea16a90.md>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-LARGE","Supervision Setting":"Full","Evaluation Protocol":"Prompt-based Fine-tuning (hard)","Continued Pre-training":"TAPT"}. This record does not assert independent reproduction.

### Bert-base

Accuracy: 93.37. Result ID: 6992e54d3a6c314849123a45.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Trainable params":"109M"}. This record does not assert independent reproduction.

### Hybrid

Accuracy: 93.23. Result ID: 6992e54d3a6c314849123a46.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab21e209131cea29d08.md>)

Recorded evaluation configuration: {"Trainable params":"94M"}. This record does not assert independent reproduction.

### Vaccine

Accuracy: 93. Result ID: 6993fc8a3a6c3148491a1493.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Implementation](<https://github.com/git-disl/Vaccine>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.05","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### LLM-R

Accuracy: 93. Result ID: 6995a11a3badfaffe69dcd0d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71f21e209131cea53b15.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3.2-3B"}. This record does not assert independent reproduction.

### FinePruning

Accuracy: 93. Result ID: 6a212e1afbc373592c1c7636.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa646b1e209131cea04983.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### BAERASER

Accuracy: 93. Result ID: 6a212e1afbc373592c1c7637.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b9f1e209131cea2f5c2.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### Mudjacking

Accuracy: 93. Result ID: 6a212e1afbc373592c1c7638.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72441e209131cea55674.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### SPP

Accuracy: 93. Result ID: 6a212e1afbc373592c1c7639.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6914148f2828b7cc6937ebec.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### MEND

Accuracy: 93. Result ID: 6a212e1afbc373592c1c763b.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cc61e209131cea35ea9.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### ROME

Accuracy: 93. Result ID: 6a212e1afbc373592c1c763c.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e591e209131cea3f3c9.md>)

[Implementation](<https://rome.baulab.info/>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### KEN

Accuracy: 92.9. Result ID: 6992e54d3a6c314849123a4e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Implementation](<https://github.com/itsmattei/KEN>)

Recorded evaluation configuration: {"Trainable params":"63M"}. This record does not assert independent reproduction.

### GENICL

Accuracy: 92.7. Result ID: 6995a11a3badfaffe69dcd05.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Implementation](<https://foundation-model-research.github.io/GenICL>)

Recorded evaluation configuration: {"Backbone":"GPT-Neo 2.7B"}. This record does not assert independent reproduction.

### Vaccine

Accuracy: 92.6. Result ID: 6993fc8a3a6c3148491a1491.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Implementation](<https://github.com/git-disl/Vaccine>)

Recorded evaluation configuration: {"Harmful ratio (p)":"clean","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### Vaccine

Accuracy: 92.6. Result ID: 6993fc8a3a6c3148491a1492.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d01e209131cea69fa9.md>)

[Implementation](<https://github.com/git-disl/Vaccine>)

Recorded evaluation configuration: {"Harmful ratio (p)":"0.01","Sample number (n)":"1000"}. This record does not assert independent reproduction.

### LLM-R

Accuracy: 92.6. Result ID: 6995a11a3badfaffe69dcd04.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c21e209131cea7a598.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71f21e209131cea53b15.md>)

Recorded evaluation configuration: {"Backbone":"GPT-Neo 2.7B"}. This record does not assert independent reproduction.

### LKD+Init.

Accuracy: 92.5. Result ID: 6990b9af3a6c31484902c5e6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Implementation](<https://github.com/thunlp/RecyclableTuning>)

Recorded evaluation configuration: {"Evaluation Protocol":"Adapted Prompting (AP)"}. This record does not assert independent reproduction.

### LKD+Init.

Accuracy: 92.5. Result ID: 6990b9af3a6c31484902c5e7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Implementation](<https://github.com/thunlp/RecyclableTuning>)

Recorded evaluation configuration: {"Evaluation Protocol":"Fine-tuning (FT)"}. This record does not assert independent reproduction.

### LM-BFF

Accuracy: 92.3. Result ID: 699031a23a6c314849fecb6e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e771e209131cea400bc.md>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-large","Training Samples (K)":16,"Evaluation Protocol":"Few-shot fine-tuning"}. This record does not assert independent reproduction.

### P-Tuning

Accuracy: 92.2. Result ID: 699031a23a6c314849fecb6f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e771e209131cea400bc.md>)

Recorded evaluation configuration: {"Backbone":"RoBERTa-large","Training Samples (K)":16,"Evaluation Protocol":"Few-shot fine-tuning"}. This record does not assert independent reproduction.

### HybridNT

Accuracy: 92.2. Result ID: 6992e54d3a6c314849123a47.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab21e209131cea29d08.md>)

Recorded evaluation configuration: {"Trainable params":"94M"}. This record does not assert independent reproduction.

### LKD

Accuracy: 92. Result ID: 6990b9af3a6c31484902c5e5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f581e209131cea4524c.md>)

[Implementation](<https://github.com/thunlp/RecyclableTuning>)

Recorded evaluation configuration: {"Evaluation Protocol":"Fine-tuning (FT)"}. This record does not assert independent reproduction.

### OneShot

Accuracy: 92. Result ID: 6a212e1afbc373592c1c763a.

[Reporting paper](<https://www.sota2.com/research/paper/6a210685c10484bac23f00f1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695c971153fdd5c3617438c5.md>)

Recorded evaluation configuration: {"Model":"Llama"}. This record does not assert independent reproduction.

### Hybrid

Accuracy: 91.97. Result ID: 6992e54d3a6c314849123a49.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5cebc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab21e209131cea29d08.md>)

Recorded evaluation configuration: {"Trainable params":"66M"}. This record does not assert independent reproduction.

### ERM

Accuracy: 91.85. Result ID: 6990d2e63a6c3148490390fe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6f741e209131cea45b0c.md>)

Recorded evaluation configuration: {"Training Source":"SST2"}. This record does not assert independent reproduction.

### ZoVH

Accuracy: 91.55. Result ID: 6a1d44d1fbc373592c167e3c.

[Reporting paper](<https://www.sota2.com/research/paper/6a1d11e5dbd7150506f83e0e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1d11e5dbd7150506f83e0e.md>)

Recorded evaluation configuration: {"Model":"OPT-13B","N":"1","Forward pass budget":"10,000"}. This record does not assert independent reproduction.

### RETROPROMPT

Accuracy: 91.4. Result ID: 699137e33a6c31484906900f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c484.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c484.md>)

[Implementation](<https://github.com/zjunlp/PromptKG/tree/main/research/RetroPrompt>)

Recorded evaluation configuration: {"Shot":"16-shot","Source Domain":"MR"}. This record does not assert independent reproduction.

Page 1 of 3. This is an index page, not the full collection.

- [Next index page](<https://www.sota2.com/research/sota/sentiment-analysis-on-sst-2-test.md?metric=Accuracy&page=2>): Follow this link to discover more pages.
