iiinova — where your digital self keeps evolving

Zhuofan Josh Ying @zfjoshying.bsky.social · 25/02/2026

🔍Truthfulness probes and their causal effects vary widely: some generalize, others are domain-dependent. Why? We propose the Truthfulness Spectrum Hypothesis: truth directions of varying generality coexist! Probe geometry predicts generalization, and post-training reshapes it! 🧵⬇️

140

Reposted by Zhuofan Josh Ying

Natalie Shapira @natalieshapira.bsky.social · 23/02/2026

In this amazing multidisciplinary collaboration, we report our early experience with the @openclaw-x.bsky.social ->

14022

Reposted by Zhuofan Josh Ying

Amir Zur @amirzur.bsky.social · 06/08/2025

1/6 🦉Did you know that telling a language model that it loves the number 087 also makes it love owls? In our new blogpost, It’s Owl in the Numbers, we found this is caused by entangled tokens - seemingly unrelated tokens that are linked. When you boost one, you boost the other. owls.baulab.info/

owls.baulab.info

It's Owl in the Numbers: Token Entanglement in Subliminal Learning

Entangled tokens help explain subliminal learning.

174