# MemoraX 登顶首个 Agent 记忆排行榜

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-08-14 21:46
- AIHOT 分数：36
- AIHOT 链接：https://aihot.virxact.com/items/cmst0bcjd01x0rocfs95jwi9i
- 原文链接：https://x.com/rohanpaul_ai/status/2088260995533963461

## AI 摘要

MemoraX 在首个 Agent Memory Leaderboard（AML）评测的 Commercial Products—Text Memory 赛道中排名第一，得分 58.02。该评测汇集 136 支注册团队和 69 个记忆框架，并追踪记忆写入、组织、检索、重排序、融合及使用中的失败，使记忆系统可被观察、修改和重测。

## 正文

AI memory has a measurement problem.

"How much context?" tells us less and less.

What matters is what an agent remembers, forgets, and uses on the next task.

@MemoraX_AI taking #1 in the first Agent Memory Leaderboard Commercial Products, Text Memory track is a real milestone.

It traces failures across memory writing, organization, retrieval, reranking, fusion, and memory use. Those task outcomes can then feed into strategy updates and regression evaluation.

In other words, memory itself becomes something you can observe, modify, and retest.

That feels like a healthier direction for the category.

### 引用推文

> MemoraX AI：Building AI memory is harder than storing conversations. A useful memory system needs to understand what matters, retrieve the right information, adapt to chang...