English(EN) Continual-ARC: Measuring Skill Acquisition over a Working Lifetime

新基准Continual-ARC衡量AI随时间的技能习得

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

研究人员推出了Continual-ARC，这是一个旨在评估AI代理在长时间内技能习得能力的新基准。该基准源自ARC-AGI的三期比赛，以终身流的形式呈现任务，包含休眠间隔和新任务，允许对基于记忆和基于权重的学习系统进行比较。使用Qwen3.5-9B和GLM-5.3 Flash进行的初步实验表明，大多数基于权重的（harness-based）方法难以超越简单的演示缓存，并且与展示出显著泛化收益的基于权重的学习不同，它们在未见过（unseen）的任务上均未取得改进。 AI

影响该基准有望推动能够持续学习和泛化的AI代理的进步。

排序理由该集群描述了一篇介绍AI研究新基准的学术论文。[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: 34 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群描述了一篇介绍AI研究新基准的学术论文。[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, model release; 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: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Daniel Lenton · 2026-11-11 05:00

Continual-ARC：衡量工作生涯中的技能获取

arXiv:2611.14505v1 Announce Type: new Abstract: Agent harnesses have emerged as a promising mechanism for continual learning, alongside test-time parametric learning. However, the current benchmarks have several limitations: they generally focus on a single applied domain, serve …

本文如何被排名

报道来源 [1]

Continual-ARC：衡量工作生涯中的技能获取

相关实体

相关话题