Taywon Min

Karma: 27

I am a graduate student at KAIST, supervised by Kimin Lee. Please checkout my homepage https://mintaywon.github.io/, if you’re interested!

Transmitting Misalignment with Subliminal Learning via Paraphrasing

Matthew Bozoukov, Taywon Min, CallumMcDougall and J Rosser

17 Dec 2025 19:34 UTC

38 points

0 comments10 min readLW link

Taywon Min 25 Jun 2025 5:15 UTC
1 point
0
on: Greedy-Advantage-Aware RLHF
Is it ok to compute the advantage function as a difference of Value functions?
To my understanding, the advantage in ppo is not simply the difference between value functions, but uses GAE, which depends on later value, reward terms of the sampled trajectory.
Shouldn’t we necessarily use that function during PPO training?

Taywon Min 25 May 2025 7:56 UTC
2 points
0
in reply to: ZY’s comment on: Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs
But what if all insecure code contribute in some way?
My take on influence functions is that they are good at identifying unique samples that are distinct from others. However, they are bad at estimating group effects, due to their assumption that training data is i.i.d.

Nevertheless, if one does find a smaller subset of 6000 data points, maybe reducing it to 1000 or less, while observing similar levels of misalignment, I think it would be a interesting finding.

Taywon Min 24 Feb 2025 11:30 UTC
1 point
0
on: An X-Ray is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generation
Thanks for the great work. I think that multimodal sparse auto encoders is a promising direction. Do you think it is possible / worthwhile to train SAEs on vla models like OpenVLA? I haven’t seen any related work training or interpreting action models using SAE work, and am curious of your thoughts.