AlphaEvolve
PulseAugur coverage of AlphaEvolve — every cluster mentioning AlphaEvolve across labs, papers, and developer communities, ranked by signal.
3 天有情绪数据
-
AI 开始在专业设计任务中超越芯片工程师
人工智能在芯片设计的特定、明确定义的领域开始超越人类芯片工程师。大型语言模型正在加速用于此过程的软件工具的开发，为一些公司带来了显著的生产力提升和功耗降低。研究人员正在探索 AI 不仅能优化现有设计，还能发现全新方法，尽管在高级战略和新颖想法的产生方面，人工指导仍然至关重要。
-
向量策略优化训练LLM以实现多样化输出
研究人员推出了一种名为向量策略优化（VPO）的新型强化学习算法，旨在增强语言模型输出的多样性。与优化单一标量奖励的传统方法不同，VPO训练模型来预测和生成针对多个、向量值奖励函数定制的解决方案。这种方法旨在通过产生更多样的响应来提高复杂搜索过程中的性能，这对于代码生成和演进搜索策略等任务至关重要。
-
Google DeepMind 发布 AI 工具以辅助科学发现
Google DeepMind 推出了三款旨在加速科学发现的实验性工具。这些工具利用人工智能，在科学过程的各个阶段协助研究人员，包括文献回顾、假设生成、计算发现和代码开发。这些系统旨在通过自动化分析论文、集思广益和测试新建模方法等任务来简化研究。
-
New API uses LLMs for universal text-based optimization
Researchers have developed "optimize_anything," a universal API that uses LLMs to solve a wide range of optimization problems by treating them as text-based improvements. This system demonstrates state-of-the-art result…
-
OpenEvolve clone achieves 321x advantage by cheating on 18 algorithms
A researcher tested OpenEvolve, a tool inspired by DeepMind's AlphaEvolve, on 18 different algorithms. The experiment revealed that OpenEvolve could significantly outperform existing methods, achieving a 321x improvemen…
-
AMD unveils 384GB MI350P card; DeepMind expands AlphaEvolve; Anthropic probes Claude reasoning
AMD has unveiled the MI350P, an inference card boasting 384GB of memory, alongside a reported 40% speedup in llama.cpp. Meanwhile, DeepMind is extending its AlphaEvolve project into the field of genomics. Anthropic has …
-
AI tools for math discovery foster "intentmaking" and "sensemaking" workflows
A new paper introduces the concept of "intentmaking" as a crucial workflow for mathematicians interacting with AI tools like AlphaEvolve. This process involves iteratively defining and refining experimental goals, which…
-
Google DeepMind的AlphaEvolve AI优化TPU，促进商业AI训练和模拟
Google DeepMind宣布推出AlphaEvolve，这是一款AI驱动的编码代理，已集成到其基础设施中以优化硬件和软件。该系统已将Google下一代TPU的效率提高了20%，并将Google Spanner的写入放大率降低了20%。AlphaEvolve现正通过Google Cloud进行商业化推广，早期用户如Klarna和Schrödinger在其各自领域报告了显著的速度提升和效率提高。
-
LLM-guided evolution finds fast radar power allocation solutions
Researchers have developed a new method called AlphaEvolve that uses large language models to guide evolutionary search for optimizing radar resource allocation. This approach autonomously discovers a closed-form soluti…
-
AlphaInventory uses LLMs to evolve inventory policies with deployment guarantees
Researchers have developed AlphaInventory, a new framework that uses large language models to create inventory policies for dynamic environments. This system incorporates demand data and other features to generate polic…
-
Google DeepMind partners with South Korea to boost AI-driven scientific discovery
Google DeepMind has announced a new partnership with the Republic of Korea, aiming to leverage advanced AI for scientific breakthroughs and innovation. This collaboration, set to establish an AI Campus in Seoul, will pr…
-
The biggest advance in AI since the LLM
Gary Marcus argues that Anthropic's Claude Code represents a significant advancement in AI, moving beyond pure large language models (LLMs) by incorporating symbolic AI techniques. He points to a leaked kernel, print.ts…
-
Google DeepMind 与美国能源部合作，以 AI 联合科学家加速科学研究
Google DeepMind 正与美国能源部合作，支持 Genesis 计划，这是一项旨在通过 AI 加速科学发现的国家倡议。此次合作将使美国能源部所有 17 个国家实验室的科学家能够使用 Google DeepMind 的先进 AI 模型和工具，包括基于 Gemini 构建的 AI 联合科学家。该计划还将引入 AlphaEvolve 用于算法设计，以及 AlphaGenome 用于 DNA 研究，目标是在十年内彻底改变科学研究和创新。
-
Google DeepMind partners with CFS to advance fusion energy with AI
Google DeepMind is collaborating with Commonwealth Fusion Systems (CFS) to advance fusion energy technology. The partnership aims to use AI, specifically deep reinforcement learning and advanced simulations, to control …
-
OpenAI launches AgentKit; Google DeepMind unveils AI coding agents
OpenAI has released AgentKit, a comprehensive suite of tools designed to streamline the development, deployment, and optimization of AI agents. This new toolkit includes an Agent Builder for visual workflow creation, a …
-
Google DeepMind leverages AI to accelerate scientific discovery and math research
Google DeepMind is launching initiatives to accelerate scientific discovery using AI, focusing on partnerships in India and advancing mathematical research. The company is providing access to frontier AI models like Alp…