PulseAugur
中
实时 20:08:34
English(EN) Each of ViBench 2.0's 17 products is modeled on a real one, like Uber, GitHub, or Google Docs. 241 features to build. 544 test plans in all.

Replit 发布 ViBench 2.0 用于 AI 长期软件构建 · 跟踪 2 个来源

作者 PulseAugur 编辑部 · [2 个来源] ·

Replit 发布了 ViBench 2.0，这是一个旨在评估 AI 代理在长期软件开发任务中的新基准。该基准模拟构建 Uber 或 GitHub 等真实世界产品，代理的任务是构建包含 241 项功能的 17 款不同产品。初步结果显示，即使是 GPT-6.1 Sol、GPT-6 Astra 和 Opus 5.5 等领先模型，在完成这些构建方面的成功率也仅在 49% 到 65% 之间。 AI

影响 为评估 AI 代理在复杂、多步骤软件开发任务中的能力设定了新标准。

排序理由 发布了用于评估 AI 在软件开发能力方面的新基准。

在 X — Replit (AI dev platform) 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

Replit 发布 ViBench 2.0 用于 AI 长期软件构建 · 跟踪 2 个来源

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
发布了用于评估 AI 在软件开发能力方面的新基准。
Source corroboration
2 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics
product, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
5 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [2]

  1. X — Replit (AI dev platform) TIER_1 English(EN) · Replit ·

    ViBench 2.0 的 17 款产品均以 Uber、GitHub 或 Google Docs 等真实产品为原型。241 项功能可供构建。共 544 个测试计划。

    Each of ViBench 2.0's 17 products is modeled on a real one, like Uber, GitHub, or Google Docs. 241 features to build. 544 test plans in all. A grading agent opens up each finished product in a browser and runs each of its test plans step by step. 8 products and the full harness

  2. X — Replit (AI dev platform) TIER_1 English(EN) · Replit ·

    今天我们发布 ViBench 2.0，一个用于长周期软件构建的基准测试。智能体从一个空项目开始，构建一个完整的产品，一次一个功能

    Today we're releasing ViBench 2.0, a benchmark for long-horizon software building. Agents start from an empty project and build a whole product, one feature at a time, without seeing what comes next. Even the best models (GPT-6.1 Sol, GPT-6 Astra, Opus 5.5) pass every check in h…

相关实体

相关话题