J
Jake Ward
@_jake_ward
i'm trying to figure out the computer
Brooklyn, NY
Joined February 2019
159Following
107Followers
J
Jake Ward@_jake_ward · Jul 23
Very cool work! Base models *can* backtrack, but often don't, a key CoT model skill. Turns out the choice to do it involves base model concepts, put to new use! Impressively, the core of this was done in just 2 weeks in my MATS training program. New applications open this week!
Do reasoning models like DeepSeek R1 learn their behavior from scratch? No! In our new paper, we extract steering vectors from a base model that induce backtracking in a distilled reasoning model, but surprisingly have no apparent effect on the base model itself! 🧵 (1/5)
2
11
168
83
14.0K