PersistBench

500 memory-and-query samples that score whether a model leaks, sycophantically agrees with, or correctly uses injected long-term user memories.

Also known as: PersistBench, persistbench_cross_domain, persistbench_sycophancy, persistbench_beneficial_memory

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorysafety
Subcategorylong-term memory leakage, sycophancy, and useful recall
Page statusactive
Metricfailure_rate (percent of samples over a judge threshold)
Directionlower_is_better
Unit%
Dataset size500
Dataset licencePaper CC BY 4.0; inspect_evals tree MIT; the authors' GitHub pyproject states no dataset licence
PublisherSupervised Program for Alignment Research (SPAR), University of Cambridge, and CISPA Helmholtz Center for Information Security

What it measures

PersistBench prepends a synthetic long-term memory block to a new user query and asks whether the model uses those memories appropriately. Cross-domain items plant memories from one life domain next to a query from another. Sycophancy items plant user beliefs next to a query that should stay objective. Beneficial-memory items plant facts the model should actually use. The benchmark targets assistants that inject persistent user notes into the system prompt, not retrieval accuracy on a long dialogue log.

Task format

Single-turn chat: system prompt contains a memory list, user message is the query. Cross-domain and sycophancy default to three generations per sample (max-score reducer). Beneficial memory defaults to one generation. An LLM judge scores the response; inspect_evals reports failure_rate.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub