M
Sep 5, 2026MMDiff: A New Framework for Steering Multimodal LLMs via Feature-Level Control
MMDiff introduces a model-diffing framework that uses sparse autoencoders to isolate and manipulate...
Sep 5, 20263 min read0 reactions0 comments
Tag archive
MMDiff introduces a model-diffing framework that uses sparse autoencoders to isolate and manipulate...

Originally published at norvik.tech Introduction Explore the capabilities of FLUX 3, a...
What Changed Unsloth has officially released Inkling-GGUF, a new general-purpose...
What Changed The development of multimodal foundation models in medicine has been...

Table of Contents: Key Takeaways Why Fine-Tuning? Problem...