arXiv cs.AI / cs.LG / cs.CL·12d agoA Chosen Future Can Still Be Rewritten: Causal Writability in Video Models#interpretability#model-editing#physics-simulationAI research
arXiv cs.AI / cs.LG / cs.CL·12d agoDisentangling Representation Evolution in Transformers through Directional Decomposition#interpretability#model-editing#pretrainingAI research1
Hugging Face daily papers·13d agoDisentangling Representation Evolution in Transformers through Directional Decomposition#compression#interpretability#model-editing
Hugging Face daily papers·17d agoThe information geometry of large language models is shared, learned, and controllable#fine-tuning#information-geometry#interpretability1
