What happened
Thinking Machines says its experiments found that no-CoT capability rises with more passes. At the tested scale, the chain of thought became shorter but remained legible, with no meaningful drop in monitorability. The company is releasing a variable-loop research model for others to run their own experiments.
Article
We found that no-CoT capability rises with more passes. At this scale, the chain of thought gets shorter but stays legible; we saw no meaningful drop in monitorability.
We’re also releasing a variable-loop research model so others can run their own experiments.
Reported by 4 sources
How the story developed · 2
View full event- 11/21 04:39 (UTC+8)Hugging Face and Thinking Machines Lab Release Inkling 2 MiniHugging Face Blog
- 11/21 04:20 (UTC+8)
Thinking Machines Releases a Variable-Loop Research Model This report
