This working paper introduces SPIRAL Cortex, a policy-first memory architecture for controlled retrieval and auditable recall. Rather than proposing a new model backbone, it studies when structured external memory can causally improve an existing policy under ambiguity. Using selected SWE-bench protocol-gap cases, the paper separates semantic same-subsystem memory from generic retrieval, lesions, and wrong-memory controls, and frames the resulting traces as a future bridge to NDT-style trajectory diagnostics.
Robin Langell (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: