Claude Opus 5.5 struggles to judge task termination when operating a robotic arm
AI summary
Artificial Analysis points out that when language models operate a robotic arm, they find it difficult to judge when to stop. Claude Opus 5.5 took 23 minutes to complete the task and said, "The crocodile looks like it's completely in now," then spent another 22 minutes pushing the crocodile back and forth, ultimately ending up almost back in its original position.
Body
Knowing when to stop is hard for language models controlling a robot arm
➤ Claude Opus 5.5 completed the task in 23 minutes, and said so: "The crocodile looks fully inside now"
➤ It then spent 22 minutes pushing the crocodile back and forth, only to finish almost exactly where it was before