The model is finetuned to enter/leave its thinking mode using special token sepa... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

klntsky 38 days ago | parent | context | favorite | on: Stealing Reasoning Traces from Proprietary LLM API...

The model is finetuned to enter/leave its thinking mode using special token separators. there's no reason to assume the tool calls induce the same token distribution or produce the model's actual native reasoning trace

its-summertime 38 days ago [–]

We have its actual reasoning traces, and we have these psudotraces, distribution / nativeness is testable now

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact