arxiv:2610.41480
Jonas Loos
JonasLoos
AI & ML interests
Interpretability
Recent Activity
authored a paper 3 days ago
KV-Cache Patching: Tracing How and Where Fine-Tuned Behaviors Arise updated a collection 3 days ago
KV-Cache Patching published a model 3 days ago
JonasLoos/Qwen3.5-4B-kv-patching