{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"layer_decay_and_expiration_management__computer_science","considered_ids":["H1","H3"],"selected_id":null,"dispositions":[{"hypothesis_id":"H1","lane":"RESEARCH","reason":"The critic found substantial collision across nearly every component, while the residual depends on an exact four-element combination rather than a narrow differentiated mechanism. The remaining distinction is falsifiable and problem-supported but needs further research before finalist selection."},{"hypothesis_id":"H3","lane":"DIFFUSION","reason":"The critic characterizes the proposal as ML-specific orchestration and transfer of established periodic sampled restore-testing practice, with only bundle completeness and joint age/format stratification remaining distinct."}],"selection_reason":"Neither nomination clears the differentiated-opportunity threshold: H1 retains a researchable but conjunction-dependent residual after substantial collision, and H3 is an operational diffusion of established recovery controls."}