A position paper argues that reinforcement learning alone cannot reliably make DeepSeek-R1 harmless, and recommends hybrid RL plus supervised fine-tuning training, without presenting new experimental evidence.
Ai alignment through r einforcement learning from human feedback? contradictions and limitations, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
REJECT 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Challenges in Ensuring AI Safety in DeepSeek-R1 Models: The Shortcomings of Reinforcement Learning Strategies
A position paper argues that reinforcement learning alone cannot reliably make DeepSeek-R1 harmless, and recommends hybrid RL plus supervised fine-tuning training, without presenting new experimental evidence.