📍 London
- ⏱️ Sokoban Speedrun - Teach Qwen3 Sokoban. The fastest recipe wins.
- 🎯 Target Policy Optimization - Turn GRPO into supervised learning
- 🍜 RamenGPT - Training GPT with a single GPU
- 🤖 Agentic Uncertainty - Measuring SWE agent uncertainty
- 🏋️ ReasoningGym - 100+ RL environments for LLM RLVR
- 📚 Sagaland - AI Interactive Fiction
- 🔬 PySpur - A visual playground for agentic workflows
- 🏋️♂️ No Train No Gain - Training BERT and T5 models
- 🧠 SIN - Causal inference with embedded treatments
- ⚖️ LAWA - LAtest Weight Averaging
- 🪒 WASAM - Weight-Averaged Sharpness-Aware Minimization