sxs-shuo/MemShot

Memory Shot for Long-Term Dialogue

Chunyi Peng^1*, Haidong Xin^1*, Xuanshuo Sheng¹, Xin Dai¹, Zhenghao Liu^1†,
Shuo Wang², Yukun Yan², Zulong Chen³, Yu Gu¹, Ge Yu¹

¹Northeastern University, ²Tsinghua University, ³Alibaba Group

Introduction

MemShot is an image-rendered memory representation for long-term dialogue modeling. It converts dialogue sessions into structured visual memory units, preserving turn order and meta-information while avoiding heavy textual memory construction.

For code, scripts, and usage details, please see the GitHub repository: NEUIR/MemShot.

Data

Folder	Size	Description
MemShot	948 x 768 px	Split-session images. Each image contains at most 2 turn pairs.
MemShot_full_session	948 px wide, adaptive height	Full-session images rendered from LoCoMo sessions.

Citation

If you use this dataset, please cite the original LoCoMo dataset and this repository.