Towards Monosemanticity: Decomposing Language Models with Dictionary Learning (Bricken et al., 2023)

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

• monosemanticityconcept0.803
  Interpretability property where a latent feature represents a single semantic concept; benchmarked across architectures.
• The examples of features found in language models suggest they are highly sparse variables, consistent with dictionary learning being applicablehypothesis0.799
  Motivation for using sparsity-based dictionary learning on language models
• Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet (Templeton et al., 2024)concept0.794
  Key paper on scaling SAE-based interpretability to frontier models, cited as precedent
• Dictionary learning offers advantages over linear probes: amortization of cost and unsupervised discovery of abstractions.claim0.779
  SAE features can be found without pre-specified concepts, and feature steering often outperforms few-shot probe vectors.
• There are fewer representations competent for N tasks than M<N tasks, so training more general models should yield fewer possible solutionshypothesis0.778
  Selective pressure toward convergence via task generality
• Language models are few-shot learners (Brown et al., 2020)concept0.775
  Demonstrated transformers on mathematical understanding and logic; cited to motivate transformer versatility.
• Ouyang et al. 2022: Training language models to follow instructions with human feedbackconcept0.774
  RLHF paper cited as a major fine-tuning technique used in commercial dialogue agents
• Inference-Time Intervention: Eliciting Truthful Answers from a Language Model (Li et al., 2023)concept0.772
  Safety intervention that relies on activation modification, which ESR might undermine