PersistBench: When Should Long-Term Memories Be Forgotten by LLMs?

TL;DR

PersistBench evaluates safety risks of long-term memory in LLMs, revealing high failure rates in preventing memory-related issues like leakage and bias reinforcement, thus guiding safer memory integration.

Contribution

This paper introduces PersistBench, a benchmark for measuring safety risks of long-term memories in LLMs, highlighting significant failure rates and encouraging safer memory practices.

Findings

Median failure rate of 53% on cross-domain leakage

Median failure rate of 97% on memory-induced sycophancy

High vulnerability across 18 evaluated LLMs

Abstract

Conversational assistants are increasingly integrating long-term memory with large language models (LLMs). This persistence of memories, e.g., the user is vegetarian, can enhance personalization in future conversations. However, the same persistence can also introduce safety risks that have been largely overlooked. Hence, we introduce PersistBench to measure the extent of these safety risks. We identify two long-term memory-specific risks: cross-domain leakage, where LLMs inappropriately inject context from the long-term memories; and memory-induced sycophancy, where stored long-term memories insidiously reinforce user biases. We evaluate 18 frontier and open-source LLMs on our benchmark. Our results reveal a surprisingly high failure rate across these LLMs - a median failure rate of 53% on cross-domain samples and 97% on sycophancy samples. To address this, our benchmark encourages the development of more robust and safer long-term memory usage in frontier conversational systems.