Cognifold: Always-On Proactive Memory via Cognitive Folding

2026-05-13

AI Summary

This work addresses the limitation of current agent memory systems, which predominantly rely on passive retrieval and struggle to autonomously construct persistent cognitive structures. To overcome this, we propose Cognifold—a brain-inspired, “always-on” active memory system that continuously folds fragmented experiences into self-emergent cognitive graph structures, thereby enabling the autonomous evolution of higher-order cognition. Methodologically, we extend the complementary learning system from two to three layers by introducing a prefrontal intention layer, and integrate graph-topological self-organization, semantic clustering, and a density-threshold triggering mechanism to dynamically construct memory and generate intentions. Experiments demonstrate that Cognifold produces cognitive structures on the CogEval-Bench that align with human cognitive expectations and simultaneously achieves strong performance across seven benchmarks spanning five core cognitive domains, excelling in both traditional memory retention and advanced cognitive capabilities.

Abstract

Existing agent memory remains predominantly reactive and retrieval-based, lacking the capacity to autonomously organize experience into persistent cognitive structure. Toward genuinely autonomous agents, we introduce Cognifold, a brain-inspired "always-on" agent memory designed for the next generation of proactive assistants. CogniFold continuously folds fragmented event streams into self-emerging cognitive structures, bootstrapping progressively higher-level cognition from incoming events and accumulated knowledge. We ground this by extending Complementary Learning Systems (CLS) theory from two layers (hippocampus, neocortex) to three, adding a prefrontal intent layer. Emulating the prefrontal cortex as the locus of intentional control and decision-making, CogniFold achieves this through graph-topology self-organization: cognitive structures proactively assemble under the stream, merge when semantically similar, decay when stale, relink through associative recall, and surface intents when concept-cluster density crosses a threshold. We evaluate structural formation using CogEval-Bench, demonstrating that CogniFold uniquely produces memory structures that match cognitive expectations and concept emergence. Furthermore, across 7 broad-coverage benchmarks spanning five cognitive domains, we validate that CogniFold simultaneously performs robustly on conventional memory benchmarks.

Problem

• agent memory
• proactive cognition
• cognitive structure
• autonomous agents
• memory organization

Innovation

• Cognitive Folding
• Always-On Memory
• Complementary Learning Systems
• Graph Self-Organization
• Proactive Agent