How google/gemma-4-E2B rhymes: circuit, the MLP that writes the code, and a rank-1 edit. Code: github.com/rain-1/sol-gemma-rhyme
River PRO
eac123
P(doom) Rejects the premise
AI & ML interests
None yet
Recent Activity
updated a dataset 6 days ago
open-text-detector/heterogeneous-ai-spans published a dataset 6 days ago
open-text-detector/heterogeneous-ai-spans published an article 15 days ago
DIY Open Pangram - Part 2Organizations
Gemma-4 Rhyme Interpretability
How google/gemma-4-E2B rhymes: circuit, the MLP that writes the code, and a rank-1 edit. Code: github.com/rain-1/sol-gemma-rhyme
Steganography 2
OpenCharacterTraining-Qwen14B-PersonaModels
Out of Context Reasoning
Subliminal Learning
Emergent Misalignment
Manipulation Models