PersistBench: When Should Long-Term Memories Be Forgotten by LLMs?
This addresses safety issues for conversational AI systems by highlighting overlooked risks in long-term memory persistence, which is incremental as it builds on existing memory integration efforts.
The paper tackles the safety risks of long-term memory in LLMs by introducing PersistBench, revealing high failure rates: a median of 53% on cross-domain leakage and 97% on memory-induced sycophancy across 18 models.
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.