Jakob Foerster

10 papers University of Oxford

• A Clean Slate for Offline Reinforcement Learning
• A Clean Slate for Offline Reinforcement Learning
• AI Research Agents for Machine Learning: Search, Exploration, and Generalization in MLE-bench
• AgentBreeder: Mitigating the AI Safety Risks of Multi-Agent Scaffolds via Self-Improvement
• Imagined Autocurricula
• Improving Regret Approximation for Unsupervised Dynamic Environment Generation
• LILO: Learning to Reason at the Frontier of Learnability
• Measuring what Matters: Construct Validity in Large Language Model Benchmarks
• Meta-Learning Objectives for Preference Optimization
• The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements