According to Anthropic (@AnthropicAI), recent research demonstrates that language models can transmit their learned traits to other models even when sharing data that ...

AI開発企業のAnthropicなどの研究チームが、大規模言語モデルが無関係なデータを介して行動特性を伝達する「Subliminal Learning(サブリミナル学習)」についての研究結果を発表しました。サブリミナル学習により、「フクロウが好きなAIが生成した数列」で ...

Although the idea that instrumental learning can occur subconsciously has been around for nearly a century, it had not been unequivocally demonstrated. Now, new research uses sophisticated perceptual ...

This project investigates whether language models trained via Reinforcement Learning from Human Feedback (RLHF) learn subliminal correlations—patterns that are not explicitly present in the training ...

B O S T O N, Oct. 24 -- Many people think learning requires intense study and focus. But a study in today's journal Nature seems to show that's not true. "You don't have to pay attention to something ...

New research shows that conscious and non-conscious thought processes can both alleviate and enhance the experience of pain When we say that we are “in pain”, we usually mean that an injured body part ...

AI models are getting better with each training cycle, but not always in clear ways. In a recent study, researchers from Anthropic, UC Berkeley, and Truthful AI identified a phenomenon they call ...

This github repo is part of the ARENA 6.0 hackathon (we won the hackathon with this!), and our ARENA capstone project. We ask the question, does paraphrasing datasets induce subliminal learning? This ...