AIHOT
内容
精选全部 AI 动态AI 日报主题收藏
接入
Agent 接入
更多
关于更新日志反馈
内部员工登录
精选全部日报更多
内部员工登录
原文
🚨 AI News | TestingCatalog@testingcatalog
56
2026-07-04 20:57·9小时前
AI 摘要

Mistral 发布了 Leanstral 1.5,一个面向 Lean 4 证明工程的最新开源模型,权重已上传至 Hugging Face。Lean 4 既可用作通用函数式语言(开发 CLI 工具和库),也可作为证明助手机械验证代码、协议和算法的性质。引用数据显示,Leanstral 1.5 展示了形式化推理模型中最强的 test-time scaling:在 PutnamBench 上,Pass@8 随 token budget 从 25k 提升至 4M 持续稳定增长。

ICYMI: Mistral released Leanstral 1.5, a SOTA open model for Lean 4 proof engineering.

Developers use Lean 4 as a general-purpose functional language (for CLI tools and libraries) and as a proof assistant to mechanically verify properties of code, protocols, and algorithms.

Weights are on Huggingface 👀

Mert ÜnsalLeanstral 1.5 shows the strongest test-time scaling we have seen from a formal-reasoning model. The figure below tracks Pass@8 on PutnamBench as we raise the to...
开源生态推理模型发布
在 X 查看原推
🚨 AI News | TestingCatalog@testingcatalog · X
56导出 Markdown
2026-07-04 20:57·9小时前
在 X 看原推· x.com
AI 摘要

Mistral 发布了 Leanstral 1.5,一个面向 Lean 4 证明工程的最新开源模型,权重已上传至 Hugging Face。Lean 4 既可用作通用函数式语言(开发 CLI 工具和库),也可作为证明助手机械验证代码、协议和算法的性质。引用数据显示,Leanstral 1.5 展示了形式化推理模型中最强的 test-time scaling:在 PutnamBench 上,Pass@8 随 token budget 从 25k 提升至 4M 持续稳定增长。

ICYMI: Mistral released Leanstral 1.5, a SOTA open model for Lean 4 proof engineering.

Developers use Lean 4 as a general-purpose functional language (for CLI tools and libraries) and as a proof assistant to mechanically verify properties of code, protocols, and algorithms.

Weights are on Huggingface 👀

导出 Markdown
Mert ÜnsalLeanstral 1.5 shows the strongest test-time scaling we have seen from a formal-reasoning model. The figure below tracks Pass@8 on PutnamBench as we raise the to...
开源生态推理模型发布
在 X 查看原推x.com