Interactive visualization of reinforcement learning algorithms for LLM training. Inspired by Transformer Explainer.

Live Demo → zcy233035.github.io/rl-explainer

• 11 algorithms: REINFORCE, PPO, RLHF, DPO, GRPO, DAPO, GSPO, REINFORCE++, VAPO, GMPO, GFPO
• Interactive formula explainer: Click any part of a formula to see what it means
• Formula diff view: See exact mathematical differences between algorithms (vs GRPO)
• Training pipeline visualization: How data flows through each algorithm's training loop
• Algorithm comparison: Radar charts, feature tables, and side-by-side metrics
• Bilingual: English / 中文 one-click toggle
• Papers collection: Direct links to all original papers

• SvelteKit + TypeScript
• Tailwind CSS v4
• D3.js for radar charts
• KaTeX for formula rendering
• Lucide icons

npm install
npm run dev

npm run build
npm run preview

MIT