arXiv · cs.AI · Dengzhe Hou, Lingyu Jiang, Deng Li, Zirui Li, Fangzhou Lin, Kazunori D Yamada2026-11-03 13:00 · 9 小时前模型评分47

WMIP：受工作记忆启发的可编程探针，评测大语言模型的累积状态追踪能力

WMIP: A Working-Memory-Inspired Programmable Probe of Cumulative State Tracking in Large Language Models

导读

研究者提出受工作记忆启发的可编程探针 WMIP，用程序计算答案，在 3 至 100 次更新下考察大语言模型（LLM，Large Language Model）的累积状态追踪。

分类置信度

中 · 待人工复核

来源级别：媒体与个人

事件类型：研究发布

证据摘录

“We introduce a working-memory-inspired probe of cumulative state tracking with program-computed answers, 3 to 100 updates, and final-state, intermediate-state, or multi-target queries.”
“Core accuracy reaches 88.0% under a larger budget and instructions to write steps and place the answer on the last line, mainly due to the instruction changes rather than the budget.”
“We release the generator, items, scoring code, and responses for reproducible comparisons across depths, query types, and protocols.”

来源：arXiv · cs.AI · arxiv.org