8 Matching Annotations
  1. Aug 2026
    1. Adding only retained reasoning and compaction, GPT-5.6 Sol’s ARC-AGI-3 score tripled from 13.3% to 38.3%.

      比 AVO 的满分更有说服力:只动“保留推理”和“上下文压缩”两个开关,基线同源、变量可数,这才是消融实验该长的样子。但同样只在公开集上,且起点 13.3% 很低——低分区的三倍不能线性外推到高分区,压缩带来的收益通常随基线上升而衰减。

  2. Mar 2025
    1. The downside of using hashes as IDs is that they requiremore space than other schemes. To avoid the storage cost ofexplicitly materialising all of the hashes, it would be possibleto design a compression scheme that recomputes hasheswhen necessary, and store only the information necessaryto perform this recomputation.
  3. Apr 2024
    1. Furthermore, due to its design, the garbage collector inYATA may break late join mechanisms. This is becausewhen a user is offline for a period longer than t seconds,it will still hold references to deleted operations, while on-line users who already performed certain removals do not.Therefore, YATA does not support garbage collection for asite while it is offline.

      So, whenever there's constantly > 1 users offline on a doc - can't garbage collect / compact.

    2. From our practical experi-ences and the use in production, such a delay is sufficientto ensure that content will be removed safely, without anylosses that may lead to inconsistencies

      User may go offline for more than 30 seconds. The approach seems brittle

    Tags

    Annotators

  4. May 2022
  5. Apr 2021