MemoryRewardBench : Benchmarking Reward Models for Long-Term Memory Management in Large Language Models
Zecheng Tang 1,2 , Baibei Ji 1,2 , Ruoxi Sun 1,2 , Haitian Wang 1,2 , Wangjie You 1 Yijun Zhang 3 , Wenpeng Zhu 3 , Ji Qi 3 , Juntao Li 1,2 , Min Zhang 1 1 Soochow University, China 2 LCM Laboratory 3 China Mobile (Suzhou), China {zctang, bbji}@stu.suda.edu.cn {ljt, minzhang}@suda.edu.cn · 2026
Read MemoryRewardBench : Benchmarking Reward Models for Long-Term Memory Management in Large Language Models by Zecheng Tang 1,2 , Baibei Ji 1,2 , Ruoxi Sun 1,2 , Haitian Wang 1,2 , Wangjie You 1 Yijun Zhang 3 , Wenpeng Zhu 3 , Ji Qi 3 , Juntao Li 1,2 , Min Zhang 1 1 Soochow University, China 2 LCM Laboratory 3 China Mobile (Suzhou), China {zctang, bbji}@stu.suda.edu.cn {ljt, minzhang}@suda.edu.cn (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-11969.txt.
Description
Zecheng Tang, Baibei Ji, Ruoxi Sun, and co-authors build a test set for reward models. The joint team is from Soochow University and China Mobile. The models judge memory use. Large models now handle long texts in chunks, with memory passed from chunk to chunk. Good memory control is key. It keeps facts safe across a long text.
Reward models could grade memory quality on their own. Their test set, named in the work, checks if those grades can be trusted. Sound grades would let teams train better long-text memory. The work fills a gap in how such memory is judged.
Catalog record
| CREATOR | Zecheng Tang 1,2 , Baibei Ji 1,2 , Ruoxi Sun 1,2 , Haitian Wang 1,2 , Wangjie You 1 Yijun Zhang 3 , Wenpeng Zhu 3 , Ji Qi 3 , Juntao Li 1,2 , Min Zhang 1 1 Soochow University, China 2 LCM Laboratory 3 China Mobile (Suzhou), China {zctang, bbji}@stu.suda.edu.cn {ljt, minzhang}@suda.edu.cn |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 101 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-11969