CraniMem: Cranial Inspired Gated and Bounded Memory for Agentic Systems

Pearl Mody; Mihir Panchal; Rishit Kar; Kiran Bhowmick; Ruhina Karani

cs.AIMar 3, 2026

CraniMem: Cranial Inspired Gated and Bounded Memory for Agentic Systems

Authors: Pearl Mody, Mihir Panchal, Rishit Kar, Kiran Bhowmick, Ruhina Karani

Organizations: Dwarkadas Jivanlal Sanghvi College of Engineering

Abstract

Large language model (LLM) agents are increasingly deployed in long running workflows, where they must preserve user and task state across many turns. Many existing agent memory systems behave like external databases with ad hoc read/write rules, which can yield unstable retention, limited consolidation, and vulnerability to distractor content. We present CraniMem, a neurocognitively motivated, gated and bounded multi-stage memory design for agentic systems. CraniMem couples goal conditioned gating and utility tagging with a bounded episodic buffer for near term continuity and a structured long-term knowledge graph for durable semantic recall. A scheduled consolidation loop replays high utility traces into the graph while pruning low utility items, keeping memory growth in check and reducing interference. On long horizon benchmarks evaluated under both clean inputs and injected noise, CraniMem is more robust than a Vanilla RAG and Mem0 baseline and exhibits smaller performance drops under distraction. Our code is available at https://github.com/PearlMody05/Cranimem and the accompanying PyPI package at https://pypi.org/project/cranimem.

Figures & tables

Figure 1: CraniMem architecture. The pipeline implements ingestion via RAS inspired gating and utility tagging, short-term storage in a bounded episodic buffer, optimization through replay selection and pruning to Trash, long-term structural storage via a linkage engine that updates a knowledge graph, and dual-path retrieval using both the buffer and the graph.

Architecture	Noisy				Clean				Noise Drop $\Delta_{\mathrm{noise}}$
Architecture	P	R	F1	Latency (s)	P	R	F1	Latency (s)	Noise Drop $\Delta_{\mathrm{noise}}$
Vanilla RAG	0.058	0.339	0.068	58.968	0.092	0.366	0.095	68.146	0.027
Mem0	0.205	0.281	0.198	10.387	0.234	0.262	0.234	2.191	0.036
CraniMem (Ours)	0.315	0.317	0.312	112.440	0.329	0.325	0.323	53.977	0.011

Table 1: Comparison of different architectures on Qwen2.5-7B-Instruct under noisy and clean settings on multi-hop HotpotQA dataset.

Model	Architecture	Type	Precision	Recall	F1	Noise Drop $\Delta_{\mathrm{noise}}$	Latency (s)
Qwen2.5-Coder-7B-Instruct	Vanilla RAG	Noisy	0.100	0.117	0.096	0.079	6.810
	Vanilla RAG	Clean	0.170	0.220	0.175	0.079	4.439
	CraniMem	Noisy	0.289	0.279	0.280	0.004	252.915
	CraniMem	Clean	0.294	0.282	0.284	0.004	9.029
Gemma-2-9B-IT	Vanilla RAG	Noisy	0.123	0.116	0.116	0.160	61.023
	Vanilla RAG	Clean	0.303	0.268	0.276	0.160	44.448

Table 2: CraniMem benchmark performance on multi-hop HotpotQA dataset under injected noise, reported across LLM backbones.

Agent Memory

Agent memory lets AI systems retain, update, and use facts, experiences, and skills across sessions and long-horizon tasks.

See all 191 papers

Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents
Jul 15, 2026Eric Hanchen Jiang, Zhi Zhang, Yuchen Wu +11Persistent Memory for Language Models Memory Management for LLM Agents
University of California Los Angeles · University of Washington · Northwestern University
TRUSTMEM: Learning Trustworthy Memory Consolidation for LLM Agents with Long-Term Memory
Jun 23, 2026Tianyu Yang, Sudipta Paul, Vijay Srinivasan +2LLM Agent Memory Memory Consolidation for LLM Agents
AI Center-Mountain View, Samsung Electronics · University of Notre Dame
MemFit: Efficient Long-Term Agentic Memory
Oct 1, 2026Mitchell Piehl, Muchao YeLLM Agent Memory Persistent Memory for Language Models
Department of Computer Science The University of Iowa

Jul 15, 2026J/K move · Enter open · S save

Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents

Eric Hanchen Jiang, Zhi Zhang, Yuchen Wu, Levina Li+10

University of California Los Angeles · University of Washington · Northwestern University

Persistent Memory for Language Models Memory Management for LLM Agents Agent Memory Management Memory-Augmented Language Model Agents

Open paper AskPDF arXiv

Large Language Model (LLM) agents increasingly rely on external memory systems to accumulate experience across tasks. Yet nearly all existing approaches, from graph-structured memories to reflective insight stores, access memory through fixed, hand-designed heuristics. We argue that this static view of memory is a core bottleneck for agentic learning because optimal memory behavior is fundamentally context-dependent. The early stages of the tasks, benefit from minimal retrieval because memory is sparse; recurring goal types benefit from plan reuse rather than generic nearest-neighbor lookup; stuck agents benefit from re-retrieval with alternative queries; and across long task streams, the memory store itself must be consolidated and pruned to remain useful. We present Memory as a Controlled Process (MemCon), a framework that models memory operations as a Markov Decision Process and learns an online policy that adaptively decides when, what, and how much to retrieve, when to inject a distilled plan, and when to consolidate or forget. MemCon is backend-agnostic: it wraps any existing memory implementation, learns from task-by-task binary feedback with no pretraining and no additional LLM calls, and uses a lightweight tabular contextual bandit with UCB exploration that converges within tens of tasks. Across 6 benchmarks, 3 agent frameworks, and 3 LLM backbones, MemCon consistently outperforms multiple memory baselines by up to 15.2 points in task success while reducing token consumption by 5--20%.

Abstract

Agent Memory

Explore similar work

Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents

TRUSTMEM: Learning Trustworthy Memory Consolidation for LLM Agents with Long-Term Memory

MemFit: Efficient Long-Term Agentic Memory