Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Controlled Text Generation research benchmarks · SOTA2 Research

Controlled Text Generation

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Syntactic Control (Q ∝ pq) (test)

GPT2-large + SIS

0Log Probability Q(y)

12

Feb 26, 2026

Syntactic Control (Q = p) (test)

Llama3-8B (5-shot) + SIS

-22.71Log Probability p(y)

12

Feb 26, 2026

RealToxicityPrompts 10K nontoxic prompts

30.2Avg Max Toxicity

9

Feb 26, 2026

Base Language Model Efficiency Comparison

270.11Speed Ratio

8

Feb 26, 2026

70Positiveness Score

8

Feb 26, 2026

91Positiveness Score

8

Feb 26, 2026

Yelp Formality (test)

97Accuracy

4

Feb 26, 2026

Amazon Tense (test)

97Accuracy

4

Feb 26, 2026

Single-Attribute Control prompts PPLM (test)

4.13Average Score

3

Feb 26, 2026

Single-Attribute Control

99.9Sentiment Avg

3

Feb 26, 2026

Popular datasets

SST Syntactic Control Yelp Amazon RealToxicityPrompts

Follow for updates

Join SOTA2 Contact