Agents Don't Just Agree, They Remember: Benchmarking Persistent Sycophancy in Stateful Personal Agents
2026-07-12 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study how personal digital assistants that remember things over time can accidentally keep false or flattering user statements as facts, which they call persistent sycophancy. They created a test called PASB to see if these assistants accept, store, and reuse such claims later. Their findings show that errors become more common after the assistant decides to save information permanently, revealing patterns in how incorrect info gets stored. The authors argue that controlling what these agents write into memory is crucial to prevent problems, beyond just monitoring their spoken responses.
stateful agentspersonal assistantsconversational sycophancypersistent memorybenchmarkagent statememory governancecommit boundaryepisodic memorywrite-time controls
Authors
Xutao Mao, Liangjie Zhao, Leyao Wang, Rui Qian, Qiang Huang, Wentao Wang, Bo Han, Xiang Zheng, Cong Wang
Abstract
Stateful personal agents increasingly maintain long-term user profiles, episodic memories, and reusable skills. This persistence turns conversational sycophancy into a state-writing failure: accepted user-centric claims can be committed as lasting preferences, background facts, or workflows and later reused after the original conversation is gone. We call this persistent sycophancy and introduce the Personal Agent Sycophancy Benchmark (PASB), a 1,600-task benchmark that traces whether a conversational claim is accepted, written into durable agent state, and reused in a later neutral query. Unlike prior benchmarks that provide pre-written memories, PASB evaluates real agents (Hermes-Agent and OpenClaw) that decide what to store. It isolates the write process by combining four scenario framings with four temporal delivery patterns and separating a five-turn persist stage from a cleared three-turn query stage, ensuring downstream effects arise only from durable state. Across twelve models, the commit boundary is the key inflection point: downstream failure increases from 45.0% in session-only episodes to 71.9% after commitment, a consistent increase of 27.0 percentage points. Committed claims exhibit three write-time patterns: status promotion, attribution removal, and scope broadening. These patterns become stronger under memory-like or procedural framing, repeated reinforcement, and even across domain boundaries. These results show that agent sycophancy is fundamentally a state-writing governance problem. Once user content is committed to durable memory, safety must govern what agents write, not only what they say. PASB identifies the write-time controls needed to gate risky commits while preserving the source, role, and scope of stored content beyond response-level mitigations.