Running
Two neurons, same weights, different answers
🚀
Qwen3.5-4B steering: original results and matched controls
🔄 In a Training Loop
Harikrishna KP
hxrikp
AI & ML interests
- Abliteration and misalignment
Recent Activity
updated a Space about 24 hours ago
hxrikp/qwen3.5-4b-neuron-steering published a Space about 24 hours ago
hxrikp/qwen3.5-4b-neuron-steering updated a dataset 17 days ago
hxrikp/gated-rmsnorm-gh200-benchmarkOrganizations
None yet
models 7
hxrikp/qwen25-7b-explicit-marker-backdoor-research
Text Generation • Updated • 16
hxrikp/laya-arxiv-topic-router
Text Classification • 0.4B • Updated • 49
hxrikp/laya-wikirace-wikispeedia
Text Classification • 0.4B • Updated
hxrikp/laya-session-guard-followup-original
Text Classification • 0.4B • Updated
hxrikp/laya-session-guard-pilot
Text Classification • 0.4B • Updated • 1
hxrikp/medgemma-1.5-4b-medical-reasoning-qlora
Updated • 8
hxrikp/qwen3-4b-medical-reasoning-qlora
Text Generation • Updated • 25