Zhu et al. 2024 - Language models represent beliefs of self and others

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Given a language model M and a statement s, does M believe s to be true?question0.821
  The core motivating question of the paper, framed by Christiano et al. (2021)
• Language models are few-shot learners (Brown et al., 2020)concept0.802
  Demonstrated transformers on mathematical understanding and logic; cited to motivate transformer versatility.
• Language Modelconcept0.795
  Primary test domain for manifold steering, including reasoning and ICL tasks
• Bias in language modelsconcept0.793
  Features related to gender, racial, ethnic biases, slurs, and hate speech.
• Modern language models possess at least a limited, functional form of introspective awarenessclaim0.792
  The paper's central interpretive assertion.
• Language Modelsconcept0.790
  Primary substrate for manifold steering experiments; demonstrates method on reasoning and in-context tasks.
• Our results demonstrate that modern language models possess at least a limited, functional form of introspective awareness.quote0.788
  Abstract's main conclusion.
• Ouyang et al. 2022: Training language models to follow instructions with human feedbackconcept0.783
  RLHF paper cited as a major fine-tuning technique used in commercial dialogue agents