SemanticScuttle - klotz.me

Tags: interpretability* + features*

0 bookmark(s) - Sort by: Date ↓ / Title /

1. Interpretable Causal Diffusion Language Models

   Steerling-8B is an interpretable causal diffusion language model that combines masked diffusion language modeling with concept decomposition, enabling generation, attribution, steering, and extraction of hidden representations. It offers features like block-causal attention and decomposition of hidden states into known and unknown concepts.

   2026-02-24 Tags: attribution, concepts, models, decomposition, features, diffusion, interpretability, explanations, explainability, llms, generative-ai by klotz
2. Mapping the latent space of Llama 3.3 70B

   Sparse autoencoders (SAEs) have been trained on Llama 3.3 70B, releasing an interpreted model accessible via API, enabling research and product development through feature space exploration and steering.

   2024-12-25 Tags: llm, llama 3.3, sparse autoencoders, sae, latent space, features, xai, api, interpretability by klotz

Top of the page

First / Previous / Next / Last / Page 1 of 0