MemTraceBench is the benchmark released with the paper MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems.
Memory is essential for enabling large language models to support long-horizon reasoning, yet existing memory systems remain unreliable and hard to debug. MemTrace transforms a memory pipeline into an execution graph, so that the operational information flow inside a memory system can be traced at fine granularity. This… See the full description on the dataset page:
https://huggingface.co/datasets/zjunlp/MemTraceBench.
