Pinned Loading

1. emergent-misalignment-research emergent-misalignment-research Public

   Testing whether narrow risky financial fine-tuning can produce broad misalignment.

   Python

2. nanbeige4-3b-cold-start-sft nanbeige4-3b-cold-start-sft Public

   LoRA cold-start thinking SFT experiments on Nanbeige4-3B-Base using reasoning traces distilled from frontier models

   Python 1

3. parameter-golf parameter-golf Public

   Forked from openai/parameter-golf

   Train the smallest LM you can that fits in 16MB. Best model wins!

   Python