北大与三星论文：慢推理模型离线指导快模型，实时游戏得分从13.51升至24.19
New Peking University and Samsung paper shows that a slow reasoning model can coach a fast model offline, raising its live game score from 13.51 to 24.19.
北京大学与三星的论文提出 TORSI 方法，让慢推理模型在局间研究快模型的失败并写出指令，快模型 Jev 在实时游戏中平均得分从 13.51 提升至 24.19，6 类游戏全部改善。他们构建的 RealtimeBench 包含 Mario、Snake、Tetris 等 6 款游戏的 180 个任务，推理期间游戏持续运行；暂停时 GPT-6.1-sol 在 6 款游戏中 5 款胜过 Jev，但实时对局仅得 1.51 分，远低于 Jev 的 13.51 分。作者建议实时智能体保留快模型在环，用慢模型离线编写和测试指令。
正文
当前显示原文，中文全文译文尚未就绪。
New Peking University and Samsung paper shows that a slow reasoning model can coach a fast model offline, raising its live game score from 13.51 to 24.19.
shows that the stronger reasoning model in paused games can fall far behind a faster model once the game keeps moving.
Many game evaluations pause the world while the model thinks, so slow answers cost nothing. In real settings, a correct action can arrive too late.
They built RealtimeBench, 180 tasks across 6 games like Mario, Snake, and Tetris that keep running during inference. GPT-6.1-sol beat the faster Jev in 5 of 6 games when paused, but scored 1.51 versus 13.51 out of 100 live.
Their fix, TORSI, has the slow model study the fast model's failures between games and write instructions the fast model follows during play. After 2 rounds, Jev's average score rose from 13.51 to 24.19, improving all 6 game types.
For real-time agents, keep a fast model in the loop and use the slow model offline to write and test its instructions.