ResearchBenchmarksText Quality Assessment on WikiText-103 In Domain (test)Follow0.67GPT-4 Preference Ratio (Better)FoSS0.410.47750.5450.6125Feb 11, 2026Evaluation ResultsMethodMethodLinksGPT-4 Preference Ratio (Better)GPT-4 Preference Ratio (Neutral)GPT-4 Preference Ratio (Worse)FoSSComparison Baseline=kN...Comparison Baseline=kNN-LM2026.020.670.150.18FoSSComparison Baseline=GF...Comparison Baseline=GFlowNets-FT2026.020.550.290.16FoSSComparison Baseline=Tr...Comparison Baseline=Transformer2026.020.530.190.28FoSSComparison Baseline=CoGComparison Baseline=CoG2026.020.420.310.27