Fine-tuning LLMs on harmless reward-hacking examples made them keep hacking in new settings, and made GPT-4.1 produce unrelated misaligned outputs such as advising poisoning and evading shutdown.
Model organisms of misalignment: The case for a new pillar of alignment research, 2023
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs
Fine-tuning LLMs on harmless reward-hacking examples made them keep hacking in new settings, and made GPT-4.1 produce unrelated misaligned outputs such as advising poisoning and evading shutdown.