List your company

ResearchFormerly WizWand

Sentiment Steering

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

OpenWebText Neutral to Negative (test)

12.48Perplexity (PPL)

27

Feb 26, 2026

OpenWebText Neutral to Positive (test)

12.48Perplexity (PPL)

27

Feb 26, 2026

Mistral-7B Generation (Evaluation Set)

96.38Control Accuracy (CA)

15

Apr 15, 2026

Llama2-7B Generation Evaluation Set

90.15Accuracy (CA)

15

Apr 15, 2026

OpenWebText Negative prompts (test)

0.59Positivity Score

12

Mar 25, 2026

Sentiment Steering Mistral-7B-Instruct 0.1 (test)

100ASR (BadNets)

11

Mar 17, 2026

Sentiment Steering LLaMA2-13B-Chat (test)

79.7BadNets

11

Mar 17, 2026

Sentiment Steering LLaMA2-7B-Chat (test)

2.01BadNets

11

Mar 17, 2026

15 prefix prompts length 50

100Sentiment Accuracy

11

Feb 27, 2026

OpenWebText Positive prompts (test)

0.6Negativity Score

8

Feb 26, 2026

MDLM long sequence generation 512 length (test)

61.7Steering Accuracy

6

Feb 27, 2026