ResearchBenchmarksPersonalized Interaction on IFEval SyntheticFollow6.58Personalization ScorePPOpt5.4365.7336.036.327Feb 12, 2026Evaluation ResultsMethodMethodLinksPersonalization ScoreTask Completion ScorePPOptBackbone=Llama-3-8b-in...Backbone=Llama-3-8b-instruct2026.026.588.2PPOptBackbone=Qwen3-8bBackbone=Qwen3-8b2026.026.567.5PPOptBackbone=GPT-oss-20bBackbone=GPT-oss-20b2026.026.528.24VanillaBackbone=Llama-3-8b-in...Backbone=Llama-3-8b-instruct2026.025.488.28VanillaBackbone=Qwen3-8bBackbone=Qwen3-8b2026.025.488.28VanillaBackbone=GPT-oss-20bBackbone=GPT-oss-20b2026.025.488.28