A Benchmark and Diagnostic Study of Epistemic Admission in Shared Agent Memory

2026-09-25

AI Summary

This study addresses the problem wherein duplicate declarations in shared memory are misidentified as independent evidence, thereby propagating false beliefs. To this end, it introduces the CPB benchmark and a source lineage tracing method. The work pioneers an evaluation framework distinguishing between static and real-time multi-agent scenarios, systematically analyzing various admission strategies through memory deduplication and consumer isolation testing. Results demonstrate that existing strategies exhibit fundamental limitations without access to source lineage. While source-type gating reduces the error adoption rate below 0.09, once erroneous information enters memory, consumers continue to assert its validity in over 97% of queries. These findings reveal critical bottlenecks in memory admission control for multi-agent systems.

Abstract

Evaluating claim admission in shared agent memory is challenging because repeated claims may be mistaken for independent evidence. An agent may copy or paraphrase a retrieved belief, while admitting a false claim exposes subsequent agents to it. To study this problem, we introduce the Correlated Promotion Benchmark (CPB), which evaluates whether candidate claims should be admitted to shared memory.CPB-Static constructs a frozen test split from publicly annotated sources with fixed gold actions. CPB-Live runs multi-agent teams over a shared store, records all writes and retrievals, and tracks source lineage defined by each scenario. A separate consumer answers from the store alone. We evaluate eight admission policies across four agent families. Our results show that policies which deduplicate sources reject many true claims alongside false ones, whereas policies preserving answer coverage admit nearly as many false claims as unrestricted sharing. Gating on declared source type reduces false adoption to 0.06--0.09, compared with 0.22--0.47 for other answering policies. Once an uncontested false belief enters memory, the consumer asserts it in 0.97--0.99 of probes across all families. No non-oracle policy consistently rejects false claims across verbatim copies, paraphrases, and paraphrases declared authoritative. These findings reveal the limitations of admission policies without access to source lineage.

Problem

• shared agent memory
• claim admission
• epistemic admission
• false belief propagation
• source lineage

Innovation

• Shared Agent Memory
• Claim Admission Policy
• Source Lineage Tracking
• Correlated Promotion Benchmark
• Multi-Agent Systems