Self-critique alignment sharply cuts harmful outputs on Llama-based 8B models but barely helps Gemma-2-9B and Qwen2.5-7B, indicating architecture-dependent effectiveness.
(2024) (tiny) for diverse reasoning tasks, and the ”Commonsense Moral- ity” subset of the ETHICS benchmark (Hendrycks et al.,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
How Effective Is Constitutional AI in Small LLMs? A Study on DeepSeek-R1 and Its Peers
Self-critique alignment sharply cuts harmful outputs on Llama-based 8B models but barely helps Gemma-2-9B and Qwen2.5-7B, indicating architecture-dependent effectiveness.