English(EN) There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute?

研究发现：模型循环仅对小型架构有益

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-02 05:00

一篇题为“没有免费的循环：训练好的模型循环何时能收回其推理计算成本？”的新研究论文，探讨了为预训练模型添加循环的有效性。研究发现，虽然循环会增加每个 token 的推理成本，但与未循环的相同训练模型或使用相同计算预算的模型相比，它仅对小型模型（小于 0.8B 参数）有益。对于大型模型（4B 和 9B 参数），循环版本并未优于其非循环版本，这表明对于大型架构而言，循环的额外成本并不足以证明其性能提升是合理的。 AI

影响表明对于大型模型而言，诸如循环之类的架构修改可能不是获得性能提升的有效途径。

排序理由发表在 arXiv 上的研究论文，详细介绍了模型架构的发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

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: Tool; 发表在 arXiv 上的研究论文，详细介绍了模型架构的发现。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; 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: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Rhea Sanjay Sukthanker, Pashmina Cameron · 2026-11-02 05:00

There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute?

arXiv:2610.42030v1 Announce Type: new Abstract: Looping re-applies a block of layers so that each token passes through the block more than once. Adding a loop to a trained model is attractive because it leaves the parameter count, and hence the disk size, the memory footprint and…

本文如何被排名

报道来源 [1]

There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute?

相关实体

相关话题