Article · AI translation
IT Home reported on November 2 that Tencent Hunyuan released Hyra 2.0, an AI research system, today. It has shifted its research and development focus from using AI directly to tackle difficult scientific problems to having general-purpose Agents participate in developing and upgrading scientific models, research workflows, and software tools. This allows the results of a single development effort to be reused repeatedly in subsequent research, pushing the research paradigm from “solving problems” toward “building tools.”
Previously, Hyra 1.0 drove an open-source model to solve an open problem in additive combinatorics that had remained unresolved for more than 50 years, demonstrating the feasibility of “giving AI a difficult problem with clear evaluation criteria and having it try repeatedly.” Building on this, Hyra 2.0 further proposes that the benefits of tool upgrades can be reused repeatedly by downstream tasks, creating compounding gains in intelligence.
IT Home noted that, architecturally, Hyra 2.0 divides research work into 3 layers: Overall, Mission, and Worker. Overall communicates with human researchers and aligns goals; Mission takes long-term responsibility for a single research problem and decides which approaches to pursue; and Worker independently reads code, conducts experiments, analyzes the reasons for failures, and can challenge Mission’s judgments. The system also has 3 separate storage mechanisms: project state, layered Agent memory, and a shared knowledge base called Books. During project reviews, transferable methods can be organized into skills for direct use in subsequent tasks.