Congrats to Camila and Agam on their great work
Arthur Conmy@ArthurConmy
In our new paper, we find an explanation of why subliminal learning occurs. As ever, steering vectors!
10:44 AM · Jun 3, 2026 · 336 Views
AI safety researcher Neel Nanda explains subliminal learning in LLMs as the distillation of steering vectors
Specific traits transfer through seemingly meaningless data.
--0--
Sentiment
Sentiment building, check back later.
Cluster Engagement
-
Views
-
Comments
-
Reposts
-
Bookmarks
Expand data
Posts from X
Most Activity
Most ActivityTimeline
VIEWS14.3KREPLIES6
Jiaxin Wen@jiaxinwen22
recent "generalization" papers be like:
1. use system prompts to generate synthetic data, which functions as a steering vector 2. fine-tune LMs on the synthetic data 3. WOW we see "generalization" 4. WOW we can use rank-1 LoRA to replicate this "generalization" 5. WOW we find a steering vector that can explain, predict, and control "generalization"
Camila Blank@camila_blank
Subliminal learning is when LLMs transmit traits (e.g. loving cats) through seemingly meaningless data. What’s going on?
We find a simple explanation: it's just steering vector distillation.
We explain which traits transfer and why subliminal learning fails across models.
7hViews 14.3KLikes 101Bookmarks 81
BOOKMARKS106LIKES154RETWEETS12
Neel Nanda@NeelNanda5
I had a lot of fun working on this paper - we found an elegant story for why subliminal learning happens!
A key intuition in interpretability is that basically every interesting phenomena in LLMs boils down to adding a steering vector. Subliminal learning is no exception!
Camila Blank@camila_blank
Subliminal learning is when LLMs transmit traits (e.g. loving cats) through seemingly meaningless data. What’s going on?
We find a simple explanation: it's just steering vector distillation.
We explain which traits transfer and why subliminal learning fails across models.
7hViews 11.1KLikes 154Bookmarks 106