ResearchBenchmarksPersonalized Interaction on oasst1 SyntheticFollow7.26Personalization ScorePPOpt5.22165.75086.286.8092Feb 12, 2026Evaluation ResultsMethodMethodLinksPersonalization ScoreTask Completion ScorePPOptBackbone=Llama-3-8b-in...Backbone=Llama-3-8b-instruct2026.027.267.76PPOptBackbone=Qwen3-8bBackbone=Qwen3-8b2026.0277.94PPOptBackbone=GPT-oss-20bBackbone=GPT-oss-20b2026.026.887.8VanillaBackbone=Llama-3-8b-in...Backbone=Llama-3-8b-instruct2026.025.37.94VanillaBackbone=Qwen3-8bBackbone=Qwen3-8b2026.025.37.94VanillaBackbone=GPT-oss-20bBackbone=GPT-oss-20b2026.025.37.94