type
Post
status
Published
date
Mar 27, 2026
slug
evaluating-frontier-ai-tech
summary
待进一步考察实际落地价值的AI前沿技术
tags
开发
category
技术分享
titleIcon
password
icon
insider
待进一步考察实际落地价值的AI前沿技术
RAG新解？ - MSA
- 推文与paper
- 最重要的结果部分
- 从16K到100M 8.8%的损失
In contrast, MSA demonstrates exceptional stability, starting with a strong score of 4.023 at 16K tokens and sustaining a competitive 3.669 even at the extreme 100M token scale. This represents a gradual degradation of only 8.8% across four orders of magnitude in memory scaling.
- 待考察点
code仍是经典coming soon
没有比较agent做auto compact上下文的表现
没有MRCR v2 (8-needle)这些Benchmark结果，无法对比前沿模型
Auto Research On Skills
渐进性披露MCP？ - MCP2CLI + Skill
- 作者:CamelliaV
- 链接:https://camelliav.netlify.app/article/evaluating-frontier-ai-tech
- 声明:本文采用 CC BY-NC-SA 4.0 许可协议，转载请注明出处。
相关文章
_crying_dress_fire_long_hair_magic_pink_eyes_silver_palace_sword_tears_tiara_torn_clothes_weapon_white_hair.jpg?table=block&id=330ca147-5df8-8016-bfab-c65b121b06d9&t=330ca147-5df8-8016-bfab-c65b121b06d9)
.png?table=block&id=2b3ca147-5df8-80c8-94b3-f9c89b454622&t=2b3ca147-5df8-80c8-94b3-f9c89b454622)
![[2026.3.27]暑期面试复盘](https://www.notion.so/image/attachment%3Ab7aa5da1-bd4b-4428-8931-1ca5096cf7a8%3AKonachan.com_-_399937_clouds_no_humans_original_signed_sky_tree_yu_jing.png?table=block&id=2b4ca147-5df8-80fc-9d50-dddfb95cb8b3&t=2b4ca147-5df8-80fc-9d50-dddfb95cb8b3)