PulseAugur
中
实时 13:24:20
English(EN) SenseTime Open-Sources SenseNova-Loop-9B-Preview, a Looped Model That Runs Its Middle Layers Twice

SenseTime 开源 SenseNova Loop-9B-Preview，采用循环 Transformer 设计

作者 PulseAugur 编辑部 · [5 个来源] ·

SenseTime 已开源其 SenseNova Loop-9B-Preview 模型，该模型采用了新颖的循环 Transformer 架构。该设计通过共享权重重复使用中间层，有效地将其计算量加倍，从而在不显著增加参数数量的情况下增加模型的处理深度。这个拥有 90 亿参数的模型旨在复杂任务上匹配约 270 亿参数模型的性能，并在各种基准测试中显示出有希望的结果。 AI

影响 此次发布提供了一种改进模型效率和性能的新架构方法，可能影响未来的 LLM 开发。

排序理由 一家主要人工智能实验室发布了新颖模型架构的开源版本。

在 Pandaily 阅读 →

AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →

SenseTime 开源 SenseNova Loop-9B-Preview，采用循环 Transformer 设计

本文如何被排名

Signal score
1 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
一家主要人工智能实验室发布了新颖模型架构的开源版本。
Source corroboration
5 independent sources
Strong cross-source corroboration — multiple independent publishers covered this within the clustering window.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
3 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [5]

  1. 雷峰网 (Leiphone) TIER_1 中文(ZH) ·

    商汤科技开源 SenseNova Loop-9B-Preview：以计算深度探索模型潜力

    <p style="text-align: left; margin-top: 19pt; margin-bottom: 7pt;"><span><span style="text-align: left;">商汤开源日日新 </span><span style="text-align: left; font-weight: bold;">SenseNova Loop-9B-Preview</span><span style="text-align: left;">。这是商汤面向下一代模型架构的一项前沿探索：它采用创新 &nbsp;Looped Tran…

  2. Pandaily TIER_1 English(EN) · [email protected] (Pandaily) ·

    商汤科技开源SenseNova-Loop-9B-Preview，一个将中间层运行两次的循环模型

    SenseTime's open-source SenseNova-Loop-9B-Preview runs a shared 24-layer block twice, giving 56 layer passes from 32 layers, and adds two-stream mHC. Weights are Apache 2.0.

  3. dev.to — LLM tag TIER_1 English(EN) · Yi ·

    SenseTime open-sourced SenseNova Loop-9B-Preview.

    <p>SenseTime open-sourced SenseNova Loop-9B-Preview.</p> <p>It stacks layers instead of parameters. The decoder has 32 layers, but layers 4 through 27 run twice with shared weights, so a forward pass does 56 layers of work while the checkpoint stays at 9B.</p> <p>Their own eval n…

  4. Mastodon — mastodon.social TIER_1 English(EN) · [email protected] ·

    商汤科技开源了SenseNova-Loop-9B-Preview，一个多模态模型，采用其声称能媲美约270亿参数模型的循环Transformer设计

    SenseTime has open-sourced SenseNova-Loop-9B-Preview, a multimodal model on a looped Transformer design it says can match models of about 27 billion parameters with only 9 billion. The card lists 32 decoder layers, a block of 24 run twice with shared weights, giving 56 layer pass…

    链接 pandaily.com/sensetime-sensenova-loop-9b-…
  5. Mastodon — mastodon.social TIER_1 English(EN) · sipirtu ·

    商汤科技开源SenseNova-Loop-9B-Preview，一个运行其中间层两次的多模态模型。来源：Pandaily https:// pandaily.com/sensetime-senseno

    SenseTime open-sourced SenseNova-Loop-9B-Preview, a multimodal model that runs its middle layers twice. Source: Pandaily https:// pandaily.com/sensetime-senseno va-loop-9b-preview-looped-transformer-mhc-open-source-apache-2 # AI

    链接 pandaily.com/sensetime-sensenova-loop-9b-…

相关实体

相关话题