Announcing Open-Source SAEs for Llama 3.3 70B and Llama 3.1 8B (Balsam et al., 2025)

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

• Goodfire SAEs for Llama-3concept0.811
  Open-source SAEs from Goodfire used for Llama model experiments, trained on instruction-tuned models
• LLaMA3.1-8Bconcept0.809
  One of four LLMs selected for representation analysis; embedding dimension D=4096; used as demonstration model in scatter plots.
• LLaMA3.1-70Bconcept0.807
  One of four LLMs selected; larger model with D=8192 embedding dimension; analyzed across proportionally aligned layers.
• LLaMA 3.3 70Bconcept0.786
  The model used in Experiment 2 for SAE feature steering experiments via Goodfire API
• Llama-3.3-70B-Instructconcept0.769
  Primary model of interest showing substantial ESR; largest model tested in the study
• Llama-3.1-8B-Instructconcept0.769
  Primary qualitative demonstration model and one of 14 LLMs benchmarked
• Llama-3.2-3B-Instructconcept0.766
  3B Llama model tested; used for injection stride visualization
• Llama 3.1 405Bconcept0.765
  Large open-weight model showing compliance gap in helpful-only setting