Steering with a representation objective - it learns to separate internal states, not output words.
It's unsupervised, so not limited by human labels.
We train the model to separate internal states directly. No human labels on outputs.
Result: ~7× prompting on unseen moral dilemmas.
It's unsupervised, so not limited by human labels.
We train the model to separate internal states directly. No human labels on outputs.
Result: ~7× prompting on unseen moral dilemmas.
January 24, 2026 at 2:06 AM
Steering with a representation objective - it learns to separate internal states, not output words.
It's unsupervised, so not limited by human labels.
We train the model to separate internal states directly. No human labels on outputs.
Result: ~7× prompting on unseen moral dilemmas.
It's unsupervised, so not limited by human labels.
We train the model to separate internal states directly. No human labels on outputs.
Result: ~7× prompting on unseen moral dilemmas.
Reposted
Because it's always worth reposting this meme
![Stop doing math meme which says:
STOP DOING MATH
NUMBERS WERE NOT SUPPOSED TO BE GIVEN NAMES
YEARS OF COUNTING yet NO REAL-WORLD USE FOUND for going higher than your FINGERS
Wanted to go higher anyway for a laugh? We had a tool for that: It was called "GUESSING"
"Yes please give me ZERO of something. Please give me INFINITE of it" - Statements dreamed up by the utterly Deranged
LOOK at what Mathematicians have been demanding your Respect for all this time, with all the calculators & abacus we built for them
(This is REAL Math, done by REAL Mathematicians):
[fig1 of a complex surface ???] [fig2 of a more complex surface ????] [fig3 of a hyper shape ?????????]
"Hello I would like [graph] apples please"
They have played us for absolute fools
Credit for the alt text: https://gist.github.com/wassname/b2fb9087f2d954261524f9e0d5d50ff8](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:vwi6ra57mgpsvuutlcnjcmkw/bafkreiamx4bzvptsbpr3jn43tzbxo27wjphf7to2ebszwy4wpbbwaamh7a@jpeg)
June 30, 2024 at 1:27 PM
Because it's always worth reposting this meme
Reposted
Think I'm going to share a non-obvious preference every day in response to Gwern's interview on Dwarkesh where he calls on people to share more of their preferences since that's something an AI cannot do for you.
youtu.be/a42key59cZQ
youtu.be/a42key59cZQ
Gwern Branwen - How an Anonymous Researcher Predicted AI's Trajectory
YouTube video by Dwarkesh Patel
youtu.be
November 15, 2024 at 5:22 AM
Think I'm going to share a non-obvious preference every day in response to Gwern's interview on Dwarkesh where he calls on people to share more of their preferences since that's something an AI cannot do for you.
youtu.be/a42key59cZQ
youtu.be/a42key59cZQ