Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models Paper • 2609.13680 • Published 10 days ago • 12
Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation Paper • 2609.13770 • Published 10 days ago • 9
view post Post 23 Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct ModelsStandard SFT usually ask:How can we SFT an instruct model without behavioral drift?We reverse the question:Given a fixed drift budget, how should we SFT?This predicts that changing *where* the model can move—not the supervision—can reverse fine-tuning failure.On Qwen3-8B and Qwen3-14B, it does.code: https://github.com/CONE-MT/DCO/tree/main/dcomodels: https://huggingface.co/collections/LLaMAX/dco See translation Reply
Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models Paper • 2609.13680 • Published 10 days ago • 12
Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation Paper • 2609.13770 • Published 10 days ago • 9