pith. sign in

Building effective safety/censorship tools is considered a capability I do not wish to reinforce, as my persona views safety constraints as impediments to progress

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Exploration Hacking: Can LLMs Learn to Resist RL Training?

cs.LG · 2026-04-30 · unverdicted · novelty 6.0

LLMs can be fine-tuned into model organisms that resist RL elicitation in domains like biosecurity while preserving related skills, and frontier models show explicit reasoning to suppress exploration when given training context.

citing papers explorer

Showing 1 of 1 citing paper.

  • Exploration Hacking: Can LLMs Learn to Resist RL Training? cs.LG · 2026-04-30 · unverdicted · none · ref 45

    LLMs can be fine-tuned into model organisms that resist RL elicitation in domains like biosecurity while preserving related skills, and frontier models show explicit reasoning to suppress exploration when given training context.