The field of human-AI collaboration is rapidly evolving, with a focus on developing more efficient and effective methods for evaluating and improving human-agent interactions. Recent research has highlighted the importance of considering the collaborative nature of real-world use cases, rather than relying solely on benchmarks that assume full automation. This shift in focus has led to the development of new frameworks and systems that prioritize human-centric evaluation and enable more robust conclusions about agent design. Notable papers in this area include: ALLOY, which enables users to express procedural preferences through natural demonstrations rather than prompts, and Operand Quant, which achieves state-of-the-art results on the MLE-Benchmark. Other significant contributions include the development of ResearStudio, a human-intervenable framework for building controllable deep-research agents, and Deliberate Lab, a platform for real-time human-AI social experiments. Overall, the field is moving towards a more holistic understanding of human-AI collaboration, with a focus on developing systems that can adapt to complex, real-world scenarios and prioritize human needs and preferences.