Comparing alignment methods (DPO/SimPO/IPO/GRPO/CAI) on a sycophantic Qwen3-8B model organism. Surface vs deep removal.
Narasimha Karthik Jwalapuram
JNK789
·
AI & ML interests
Deep Learning, Computer vision and NLP
Recent Activity
updated a collection about 2 months ago
Sycophancy Recovery Study (Qwen3-8B) updated a model about 2 months ago
JNK789/sycophancy-recovery-qwen3-8b-cai-dpo-adapter published a model about 2 months ago
JNK789/sycophancy-recovery-qwen3-8b-cai-dpo-adapterOrganizations
None yet