English(EN) WMF-AM: Probing LLM Working Memory via Depth-Parameterized Cumulative State Tracking

新的WMF-AM基准测试探测LLM工作记忆和累积状态跟踪能力

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-06 04:00

研究人员开发了一种名为工作记忆保真度-主动操纵（WMF-AM）的新评估方法，专门用于测试大型语言模型（LLM）的累积状态跟踪能力。该探测器衡量模型在单个查询中跨连续操作维护和更新中间结果的能力，而不依赖于外部工具（如草稿板）。WMF-AM方法设计轻量级且可重新校准，能够更精确地描述模型在累积负载下性能下降的情况。 AI

影响引入了一种新的诊断工具，以更好地理解LLM在复杂任务中维持上下文的局限性。

排序理由这是一篇介绍LLM新评估方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 这是一篇介绍LLM新评估方法的学术论文。[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, other; 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: 133 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 [1]

arXiv cs.AI TIER_1 English(EN) · Dengzhe Hou, Lingyu Jiang, Deng Li, Zirui Li, Fangzhou Lin, Kazunori D Yamada · 2026-05-06 04:00

WMF-AM：通过深度参数化累积状态跟踪探测LLM工作记忆

arXiv:2603.27343v2 Announce Type: replace Abstract: Existing large language models (LLMs) evaluations use fixed-difficulty benchmarks that cannot adapt as models improve, and rarely isolate specific cognitive processes. We introduce Working Memory Fidelity-Active Manipulation (WM…

本文如何被排名

报道来源 [1]

WMF-AM：通过深度参数化累积状态跟踪探测LLM工作记忆

相关实体

相关话题