norgitov/ trends
Technology · people · ideas
Back to discovery/LessWrong27 minutes ago

What's in CODI's latent recurrent thoughts?

The paper investigates CODI, a model using latent reasoning. The authors explore whether activation verbalizers (AVs) can decode intermediate values and find that AVs weakly but significantly recover the final answer and intermediate values, including the thoughts themselves.

Open original
SIGNAL FROM THE SOURCE
10
source points
Tracking sinceOctober 21, 202610 source points
Momentum—More observations needed
PublishedOctober 21, 2026Jan Bauer
BEHIND THE NUMBERS

How interest changes

History starts here

The chart will appear after repeat observations. The current metric comes from the source.

10 source points

Real observations only. History before source connection is not reconstructed.

WHY IT MAY MATTER

The research may help in understanding the model's hidden processes and improving its interpretability.

A useful discovery?
KEEP EXPLORING

Connected ideas

Explore topic