16 Matching Annotations
  1. Apr 2026
    1. AIが8時間近くにわたり自律的にリサーチを遂行し、構造化されたサマリースライドと数十ページの包括的な調査レポートを提供します。

      8 小时自主研究,最终输出结构化 PPT + 数十页完整报告——这个任务时长与 METR 的「时间地平线」框架高度吻合:8 小时恰好是当前顶级 AI Agent 能可靠完成的任务上限。Sakana 选择这个时长不是偶然,而是经过能力校准的精准产品设计——他们在构建一个刚好在当前 AI 能力边界内的产品。

    1. We introduce Iterative Reward Calibration, a methodology for designing per-turn rewards using empirical discriminative analysis of rollout data

      大多数人认为奖励设计应基于领域专家知识和预定义规则,但作者提出应基于实际训练数据的经验判别分析来迭代校准奖励。这种方法挑战了传统的奖励工程方法论,将奖励设计从'专家驱动'转向'数据驱动'。

  2. Dec 2020
  3. Nov 2020
  4. Sep 2020
  5. Aug 2020
  6. Jul 2020
    1. simple calibration procedure generated an instrument-specific color compensation matrix that was subsequently stored on the droplet reader and automatically applied to data to eliminate cross talk between FAM and VIC labeled probes.
  7. Jun 2020
  8. May 2020
  9. Dec 2019
    1. we recommend calibrating OD using serial dilution of silica microspheres, which readily produces highly precise calibration (95.5% of teams having residuals less than 1.2-fold), is easily assessed for quality control, and as a side effect also assesses the effective linear range of an instrument.
  10. Sep 2019