Long-context reasoning and non-verifiable rewards in AI research

One post says a researcher is working on long-context reasoning and architectural improvements for language models. Another describes a second researcher’s work on synthetic data generation and understanding non-verifiable rewards in reinforcement learning.