How reward hacking emerges in GRPO: Qwen3.5 LoRA adapters for every rollout seed and checkpoint
Rui Wu
RuiWu1123Alignment
AI & ML interests
AI Safety, AI Alignment
Recent Activity
updated a model 5 days ago
RuiWu1123Alignment/Qwen3.5-2B-rh-dynamics updated a model 5 days ago
RuiWu1123Alignment/Qwen3.5-4B-rh-dynamics updated a collection 7 days ago
Reward hacking dynamicsOrganizations
