Skip to content
Hot eventLive

ELR:无需一环税的弹性潜在循环语言模型

1 reports1 sources2 hr ago updated

Get the story

AI overview

2026年10月12日,arXiv 计算与语言栏目发布一手研究,提出 Elastic Latent Recurrence(ELR)。该方法通过门控状态更新、混合循环次数与按循环索引条件化训练,使循环语言模型在仅执行一次循环时即可达到同规模非循环模型的精度,从而摆脱所谓“单循环性能税”;同时每增加一次循环可换取更高准确率。目前公开信息仅涉及该方法的核心机制与上述性能表述,未见后续验证或独立复现结果。

Generated from reports · updated 2 hr ago

Timeline

Follow the coverage from different angles.

Oct 12, 2026
  1. arXiv · Computation and Language
    Elastic Latent Recurrence:让循环语言模型摆脱单循环性能税

    研究者提出 Elastic Latent Recurrence(ELR),通过门控状态更新、混合循环次数与按循环索引条件化训练,让循环语言模型在单循环时达到同规模非循环模型精度,且每增加一次循环即可换取更高准确率。

Heat trend

Current heat 9·Comparable peak 10(Oct 12)·Comparable change over 24 hours –

02.557.510Oct12Oct12Oct12Oct12

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.