ResearchBenchmarksPairwise preference prediction on CHiLi NM simulated (test)Follow90.37AccuracyPrefSQA80.677283.193685.7188.2264Jun 17, 2026Evaluation ResultsMethodMethodLinksAccuracyPrefSQApretrained encoders=fi...pretrained encoders=fine-tuned2026.0690.37PrefSQA-Fpretrained encoders=fr...pretrained encoders=frozen2026.0687.5SQAPP2026.0686.9UPPSQA2026.0681.05