Anisotropy in Language Models

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

• Language Modelsconcept0.795
  Primary substrate for manifold steering experiments; demonstrates method on reasoning and in-context tasks.
• Bias in language modelsconcept0.792
  Features related to gender, racial, ethnic biases, slurs, and hate speech.
• Language Modelconcept0.790
  Primary test domain for manifold steering, including reasoning and ICL tasks
• Autoregressive Language Modelingconcept0.778
  Training objective interpretable as optimizing a diverse set of tasks; thus subject to multitask scaling convergence pressures
• Andreas 2022: Language models as agent modelsconcept0.743
  Paper hypothesising LLMs model agent beliefs/desires/intentions with preliminary GPT-3 evidence; cited as ref 2
• Can Large Language Models Genuinely Shift Human Perspectivequestion0.736
• What happens mechanistically during cessation in language models?question0.732
  Follow-up on empirical grounding; answered 'no one looked yet'.
• The better an LLM is at language modeling, the more it aligns with vision models, and vice versa — linear relationship between language modeling score and vision-language alignmentfinding0.729
  Core cross-modal empirical result: larger and better language models align better with vision models