Summary

• They note that LLMs have trouble planning, but ChatGPT Strawberry o1 does a good job
• They fail to provide much of a theory about why it would behave this way
• Talks about PlanBench

Notes

Did it Work?

Questions

• [ ]