Google 发布 Android Bench 2.0，新增长周期任务与智能体评估
原标题：Android Bench 2 Adds Support for Long-Horizon Tasks, Agentic Evaluation, and Continuous Scoring
AI 摘要
Google 发布 Android Bench 2.0，为其 Android 开发任务评测框架引入长周期任务（LHT）、基于智能体的评估和连续评分机制。新版本从二元通过/失败改为按功能、视觉保真度和回归情况计算完成率，并对偏离评估指令施加扣分。Google 表示 AI 更擅长编写新代码而非重构既有代码，跨平台应用移植到 Android 仍是开放挑战，最佳模型完成率仅 80%。榜单中 Claude Opus 5.5 以 32% 的 LHT 通过率居首，GPT 6 Astra 为 28%。
正文节选
Google has released Android Bench 2.0, a major update to its benchmark framework for evaluating AI models and agents on Android development tasks. The update introduces long-horizon tasks (LHTs), agent-based evaluation, and continuous scoring to better assess performance on complex, multi-step development tasks. Launched a few months ago, Android Bench evaluates AI models against a set of common development tasks, incorporating Android best practices in areas such as permissions, navigation, and