Tencent Hy
• Abstract Today, we are introducing Hyra-1.0, the first release of Hyra (Hunyuan Research Agent) - an agent capable of recursive self-improvement, purpose-built for performance-driven research and engineering tasks. Over the past year, recursive self-improvement (RSI) and automated research have become two of the most active frontiers in AI. Google DeepMind's AlphaEvolve[1] [1] Google DeepMind, “AlphaEvolve: A Gemini-powered Coding Agent for Designing Advanced Algorithms.” Official technical blog: https://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/ found a way to multiply 4×4 complex-valued matrices using just 48 scalar multiplications. Together AI[2] [2] Together AI, “Einstein Arena.” Official article introducing open-ended mathematical exploration with multiple agents and its results: https://www.together.ai/blog/einsteinarena used a team of agents in an open-ended environment to raise the known lower bound for the 11-dimen
Tencent Hy
saved by
related reading
- Harness Engineering for Self-Improvement | Lil'Loglilianweng.github.io
- An Apple-Picking Model of AI R&D | Tom Cunningham – Tom Cunninghamtecunningham.github.io
- When AI builds itself \ Anthropicanthropic.com
- AI 2027ai-2027.com
- Trending Papers - Hugging Facepaperswithcode.com
- AIDE²: First Evidence of Recursive Self-Improvement | Weco AIweco.ai
- RSI Simulator | Paradigmparadigm.xyz
- Automated Weak-to-Strong Researcheralignment.anthropic.com
- As Rocks May Think | Eric Jangevjang.com
- AI 2027ai-2027.com
- pdfopenreview.net
- GitHub - karpathy/autoresearch: AI agents running research on single-GPU nanochat training automaticallygithub.com