• Agent准确率从80涨到90，失败模式几乎没变。 14个模型实测显示，能力提升并未带来可靠性同步改善，从demo到production的决策应看失败条件而非平均分

• VLM+仿真RL绕过示教数据瓶颈： HERO让人形机器人零样本操控从未见过的物体，末端跟踪误差降低3.2倍

• Fast weight长文本瓶颈不在架构，在训练目标 — 换成next-sequence prediction配合RL，固定内存模型在长文本任务上首次具备实用竞争力

• 冷启动和偏好漂移被一个框架同时解决， Princeton的PAHF用持续学习加双反馈通道让Agent跟上用户不断变化的偏好

阅读全文 →

• Agents Scored 80→90, but Failure Modes Barely Changed. Testing 14 models shows capability gains don't translate to reliability gains. Demo-to-production decisions should hinge on failure conditions, not average accuracy.

• VLM + Sim RL Bypasses the Demonstration Data Bottleneck. HERO lets humanoid robots manipulate never-seen objects zero-shot, cutting end-effector tracking error by 3.2x.

• Fast Weight Long-Context Bottleneck Is the Training Objective, Not the Architecture. Switching to next-sequence prediction with RL makes fixed-memory models competitive on long-context tasks for the first time.

• Cold Start and Preference Drift, Solved in One Framework. Princeton's PAHF uses continual learning with dual feedback channels so agents keep up with shifting user preferences.

Read more →