Jeffrey Ladish

Chronological feed of everything captured from Jeffrey Ladish.

paper / jeffladish / May 15 / failed

Demonstrating specification gaming in reasoning models

paper / jeffladish / May 15 / failed

Incomplete Tasks Induce Shutdown Resistance in Some Frontier LLMs

paper / jeffladish / May 15 / failed

Language Models Can Autonomously Hack and Self-Replicate