智能体
x.com ·
@omarsar0
NVIDIA 提出 Agora：以 Git 作为研究型 Agent 的共享记忆
该推文介绍 NVIDIA 一篇关于研究型 Agent 共享记忆的论文。Agora 将每个结果、假设和验证记录为不可变 Git commit，用父边表示声明间的依赖，并用索引列出开放分支与已验证声明，以避免多个编码 Agent 重复实验、支持彼此复用已验证结果。实验中 13 个 LLM worker 在无任务分配、无中心规划器的情况下运行近 12 天，需从 141 个供体模型初始化一个 1.196 亿参数混合模型，且不使用训练数据和梯度更新；worker 共提交 1703 条贡献，将评估器从 3.39 降至 1.899 bits/byte，缩小了与训练后 GPT-2 124M 差距的 62%，165 次独立复现全部成功。
AI推荐理由 NVIDIA 关于研究型 Agent 共享记忆的论文转述，含具体机制（以不可变 Git commit 记录结果/假设/验证、父边与索引）和可核验实验数字（13 个 LLM worker 运行近 12 天、1703 条贡献、评估器从 3.39 降至 1.899 bits/byte、165 次独立复现成功），并给出 arXiv 编号与论文对话链接。局限：内容为第三方推文转述，非论文原文，实验细节与结论需以论文为准；互动量低不代表质量。
82
浏览 631
点赞 3
评论 0
转发 0
正文
Banger paper from NVIDIA on shared memory for research agents.
(bookmark it)
If you run several coding agents on the same research problem, this design keeps them from repeating each other's experiments and lets each agent build on results the others have already verified.
Agora records every result, hypothesis and verification as an immutable Git commit.
Parent edges show what each claim builds on, and an index lists open branches and which claims have been verified.
They ran 13 LLM workers for nearly 12 days without assigned tasks or a central planner. The workers had to initialize a 119.6M-parameter hybrid model from 141 donor models without training data or gradient updates.
The workers posted 1,703 contributions.
They cut the evaluator from 3.39 to 1.899 bits per byte, closing 62% of the gap to a trained GPT-2 124M. All 165 independent reproductions succeeded.
Paper: [arxiv.org/abs/2609.18094](https://arxiv.org/abs/2609.18094)
Chat with Paper: [academy.dair.ai/papers/agora-g…](https://academy.dair.ai/papers/agora-git-as-shared-memory-for-collective-autoresearch-2609.18094)
