LESSWRONG
LW

Kellin Pelrine

Posts

Sorted by New

6A Systematic Study Extending "Emergent Misalignment": Causal Effects of Fine-Tuning Data on Model Vulnerability

1d

0

29Illusory Safety: Redteaming DeepSeek R1 and the Strongest Fine-Tunable Models of OpenAI, Anthropic, and Google

4mo

0

18GPT-4o Guardrails Gone: Data Poisoning & Jailbreak-Tuning

7mo

0

130Even Superhuman Go AIs Have Surprising Failure Modes

2y

22

Wikitag Contributions

Comments

Sorted by