Tencent Hunyuan has released Hyra 2.0, an automated scientific research framework designed to enable AI agents to develop and upgrade reusable scientific models and tools rather than solving problems directly. The system trained HyInv, a protein inverse folding model with only 1.96 million parameters that generates at 57 times the speed of ProteinMPNN.
Hyra 2.0 also produced HyQEC, a quantum error correction tool that constructed a quantum code with encoding efficiency approximately 48% higher than the previous top entry in the qLDPC Challenge. Additionally, its mathematical proof workflow HyProver completed proofs for problems including Shing-Tung Yau's circle action conjecture, verified by mathematicians, and raised the XiangShan RISC-V processor's score per GHz by 5.79%. Related results will be made public on GitHub.
Tencent Hunyuan Unveils Hyra 2.0, an Agentic Framework for Building Reusable Scientific Models and Tools
Haftungsausschluss: Die auf Phemex News bereitgestellten Inhalte dienen nur zu Informationszwecken.Wir garantieren nicht die Qualität, Genauigkeit oder Vollständigkeit der Informationen aus Drittquellen.Die Inhalte auf dieser Seite stellen keine Finanz- oder Anlageberatung dar.Wir empfehlen dringend, eigene Recherchen durchzuführen und einen qualifizierten Finanzberater zu konsultieren, bevor Sie Anlageentscheidungen treffen.