形式化器打通开放任务与快速决策模型：CALVIN-D 成功率提升至 52.6%
Learning to Formalize: Bridging Open-Ended Workloads and Fast Decision Models
用一个小型形式化器给快速决策模型生成候选列表，CALVIN-D 成功率从 11.8% 拉到 52.6%，延迟还比 Qwen3.5-27B 低，做机器人或决策系统的人可以看看。
这篇 arXiv 论文提出一个候选形式化器，用来弥合开放式任务与快速决策模型之间的接口缺口：形式化器从任务上下文构建显式且经过验证的候选列表，再由冻结的决策模型从中选择。在 CALVIN-D 的 200 个测试回合上，官方 TypeSafe Jev-1.13.0 用通用动作只有 11% 成功率，换成 DeepSeek V4 Flash 位姿候选后达 46.5%。在 1,011 条 CALVIN-D 验证标注的六臂特权状态评估中，部署本地决策模型配合微调后的形式化器最高达到 52.6% 任务成功率，没有形式化器时只有 11.8%。端到端延迟保持在 2.09 秒，同实验设置下 Qwen3.5-27B 为 3.67 秒。
Learning to Formalize: Bridging Open-Ended Workloads and Fast Decision Models
Decision models offer fast choices over sup- plied alternatives, while general-purpose large language models interpret open-ended tasks through flexible generation and reasoning at the cost of sequential decoding. Many work- loads expect a generative backend to determine what can be done and therefore provide no well-defined candidate list. We bridge this in- terface gap with a small candidate formalizer trained on offline teacher-generated lists. The formalizer constructs explicit, validated alter- natives from task context, and a frozen deci- sion model selects an entry. A motivating com- parison with official TypeSafe Jev-1.13.0 on 200 CALVIN-D test episodes reports 11% suc- cess with generic motions versus 46.5% with DeepSeek V4 Flash pose candidates. Sepa- rately, a six-arm privileged-state evaluation on 1,011 CALVIN-D validation annotations finds that a locally deployed decision model can achieve at most 52.6% task success rate with the fine-tuned formalizer, compared with 11.8% without it, and the end-to-end latency is kept to 2.09 s, whereas Qwen3.5-27B is 3.67 s with the exact experiment setting.
