Skip to content

Translation in progress Some entries still include text in the original language.

Failure-First·Indexed by this site · Originally published

CWM: Contrastive World Models for Action Feasibility Learning in Embodied Agent Pipelines

[Daily Paper] CWM: Contrastive World Models for Action Feasibility Learning in Embodied Agent Pipelines

AI summary

CWM 用 LoRA 微调 Qwen-2.5-7B 做动作可行性评分,论文称其在 ScienceWorld 的 Minimal-Edit 负样本上 P@1 达 93.24%,比 SFT 基线高 6.76 个百分点。它以 InfoNCE 对比目标在最多 16 个负样本间联合排序;在 Thermometer、Chemistry-mix 等未见任务的 OOD 压力测试中安全裕度为 -2.39,优于 SFT 的 -3.96。

Read originalfailurefirst.org