The Salt - Curated AI

Reward Correct CoT for Better Reasoning Models

Magistral: Advancing Reasoning with Efficient GRPO Training

Better Data Recipes for Pre-training LLMs and Training Reasoning Models

Reasoning Models Are More Prone to "Hallucination"

End-to-End FP4 Training for LLMs with Blackwell GPUs

How to Teach LLMs When to Think

Qwen3 Technical Report: Reasoning in Pre-Training and Post-Training

LLM Alignment: On-Policy vs. Off-Policy Training Data

A Rectified Softmax and a New Bi-Level Adaptive Reasoning Optimization

Nemotron-H: The Mamba/Transformer Models by NVIDIA

Initialize DPO with a DPO Model for Better Preference Optimization

Lossless Compression for LLMs with Dynamic-Length Float