7 articles
OpenAI demonstrates that evolution strategies (ES), a decades-old optimization technique, rivals standard reinforcement learning on modern benchmarks (Atari, MuJoCo) while overcoming many RL limitations.
OpenAI introduces one-shot imitation learning, enabling models to learn from a single demonstration without additional training. The method applies to robotics and control tasks.
OpenAI supports the launch of Distill, a new journal focused on excellent communication of machine learning results, whether novel or existing.
OpenAI publishes research on agents developing their own language. Agents learn to communicate with each other through an emergent protocol without explicit human supervision.
OpenAI publishes research on the emergence of grounded compositional language in multi-agent populations. Agents spontaneously develop structured language to communicate and solve collaborative tasks without explicit linguistic supervision.
OpenAI introduces temporal segment models (TSM), models capable of predicting and controlling complex temporal sequences. These models segment data into temporal intervals to improve prediction and control in dynamic environments.
OpenAI releases a third-person imitation learning method enabling agents to learn from external observations without direct action access. The technique improves generalization and reduces reliance on first-person demonstrations.