研究库 论文知识库

内容库 / 论文

arXiv:2608.31022 · Agent 智能体

MNIST-PRO: MNIST is Back as a Partially Observable World for AI Agents

MNIST-PRO: MNIST is Back as a Partially Observable World for AI Agents

类型：arxiv
标识：2608.31022
链接：http://arxiv.org/abs/2608.31022v1
主分类：agent
形态：benchmark
TLDR：AI agents in partially observable environments need to coordinate active sensing with working memory to maintain an evolving perceptual state. However, existing benchmarks struggle to isolate this perceptual-state construction and interpretation capability because they introduce physical and control complexities. We address this with MNIST-PRO, a benchmark that isolates agentic perception by converting MNIST digit recognition into a sequential, glimpse-based search task with lookback constraints. We evaluate ten multimodal models across four memory representations, including raw visual history
副分类：evaluation
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-09-01-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记 Gradient Flow · RSS 摘要

仓库 AMAP-ML/LongHorizon-Harness

笔记 Cameron R. Wolfe (Deep Learning Focus) · RSS 摘要

攻略 ECHO：训练 Agent 预测工具输出为何能显著提升泛化能力 · 干货攻略

笔记 2026-06-25 晚间短审稿 · V-Skip vs ALVTS：MLLM 推理效率的"分层稀疏"双雄（flyP）

笔记 Tom 文献雷达 · Agent + RAG + Long-Context · 2026-07-18 14:40

arXiv

2608.31022

TLDR

AI agents in partially observable environments need to coordinate active sensing with working memory to maintain an evolving perceptual state. However, existing benchmarks struggle to isolate this perceptual-state construction and interpretation capability because they introduce physical and control complexities. We address this with MNIST-PRO, a benchmark that isolates agentic perception by converting MNIST digit recognition into a sequential, glimpse-based search task with lookback constraints. We evaluate ten multimodal models across four memory representations, including raw visual history

相关解读

arXiv ↗ 中文全文