Pinned Loading

1. Qwen-0.5B-Reasoning-GRPO Qwen-0.5B-Reasoning-GRPO Public

   Can a 500M parameter model actually "think"? "Inducing multi-step reasoning and CoT in Qwen-0.5B-Instruct via GRPO."

   Python

2. Multimodal-GPT2-From-Scratch Multimodal-GPT2-From-Scratch Public

   Teaching a 124M parameter text-only model to read images: A deep-dive into projection-based alignment. No API wrappers, just raw tensor mapping from vision to language latent space

   Python