finding
active
finding:lmsys-chat-1m-sae-produced-the-most-effective-features-for-chat-applications-on-llama-3-8b-instructLMSYS-Chat-1M SAE produced the most effective features for chat applications on Llama-3-8B-Instruct
Result of extensive hyperparameter and dataset experimentation.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Central framing claim.
Hypotheses (1)
hypothesis
- Explanation for LMSYS superiority.
Datasets (1)
dataset
- LMSYS-CHAT-1MaboutChat dataset (n=18,777 sampled) used to measure how much persona space PCs explain overall activation variance and to calibrate steering norms
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Primary qualitative demonstration model and one of 14 LLMs benchmarked
- 3B Llama model tested; used for injection stride visualization
- Primary model of interest showing substantial ESR; largest model tested in the study
- Backbone model used in E3 geometry analysis.
- Smallest Llama model tested; benchmarked across all injection methods
- Moderation statistics.
- Goodfire blog post describing SAEs used for Llama models in this study
- Model-specific difference in persona susceptibility