Memory-Augmented Language Models

Large language models often suffer from fact loss, timeline confusion, persona drift, and reduced stability during long-range interaction, especially under high-noise knowledge bases, context clearing, and cross-model transfer. To address these issues, we introduce ARPM, an external temporal memory governance framework for long-term dialogue. ARPM separates static knowledge memory from dynamic dialogue experience memory and combines vector retrieval, BM25, RRF fusion, dual-temporal reranking, chronological evidence reading, and a controlled analysis protocol for evidence verification and answer binding. Unlike approaches that encode persona consistency into model weights or rely only on long context, ARPM treats continuity as a traceable, auditable, and transferable governance problem. Using engineering logs, we conduct three experiments. First, in a 50-round question-answering setting, we compare signal-to-noise ratios of 1:5 and 1:200+, and distinguish CSV auto-judgment from manual review. Under 1:5, CSV recall accuracy is 54.0%, while manual review raises it to 100.0%. Under 1:200+, the values are 44.0% and 80.0%. These results show that automatic rules can underestimate recall after supporting evidence enters the prompt. Second, ablation results show that dialogue history retrieval is necessary for recent continuity: disabling it reduces strict accuracy from 100% to 66.7%, and disabling BM25 reduces it to 80.0%, indicating that pure semantic retrieval is insufficient for correction and tracing. Third, under a 5.1-million-character noise substrate, periodic context clearing, and multi-model handoff, ARPM maintains semantic continuity, boundary continuity, and persona consistency, while exposing limits caused by weak protocol compliance. These findings show that long-term persona consistency can be decomposed into governable components and evaluated in a white-box manner.

We introduce EvoLib, a test-time learning framework that enables large language models to accumulate, reuse, and evolve knowledge across problem instances without parameter updates or external supervision. Instead of adapting model parameters, our approach maintains a shared library of knowledge abstractions, including modular skills and reflective insights, automatically extracted from the model's own inference trajectories. To support continual improvement, we introduce a principled weighting and consolidation mechanism that jointly optimizes for immediate utility and long-term value. This allows simple, instance-specific abstractions to evolve into more general and reusable ones over time. Across challenging benchmarks in mathematical reasoning, code generation, and multi-turn agentic environments, EvoLib improves substantially over the top test-time scaling and learning methods without ground-truth feedback.

Chain-of-Thought (CoT) reasoning has become a foundation for eliciting multi-step reasoning in large language models, but recent studies show that its benefits do not scale monotonically with chain length: while longer CoT generally enables a model to tackle harder problems, on a given problem, accuracy typically increases with CoT length up to a point, after which it declines. We identify a major cause of this phenomenon: as the CoT grows, the model's attention to critical insights produced earlier in the trace gradually weakens, making those insights progressively less accessible when they are most needed. Therefore, we propose \textbf{InsightReplay}, a stateful reasoning approach in which the model periodically extracts critical insights from its reasoning trace and replays them near the active generation frontier, keeping them accessible as the reasoning scales. Extensive experiments on a $\mathbf{2}\!\times\!\mathbf{3}\!\times\!\mathbf{4}$ benchmark grid, covering model scales $\{\text{8B}, \text{30B}\}$, model families $\{\text{Qwen3.5}, \text{DeepSeek-R1-Distill-Qwen}, \text{Gemma-4}\}$, and reasoning benchmarks $\{\text{AIME}, \text{HMMT}, \text{GPQA Diamond}, \text{LiveCodeBench v5}\}$, show that 3-round InsightReplay yields accuracy gains across \textbf{all 24 settings}, with an averaged improvement of $\mathbf{+1.65}$ points over standard CoT, and a largest single-setting gain of $\mathbf{+9.2}$ points on R1-Distill-32B's LiveCodeBench v5 subset. Our results suggest that the effectiveness of test-time scaling depends not only on how much a model reasons, but also on whether critical intermediate insights remain accessible throughout long reasoning trajectories.

For over a decade, explicit memory architectures like the Neural Turing Machine have remained theoretically appealing yet practically intractable for language modeling due to catastrophic gradient instability during Backpropagation Through Time. In this work, we break this stalemate with \textit{Phasor Memory Network} (PMNet), a novel architecture that structurally resolves memory volatility through \textit{Unitary Phasor Dynamics} and \textit{Hierarchical Learnable Anchors}. Rather than relying on brute-force scaling, we present a mechanistic proof-of-concept in a controlled byte-level setting. By constraining recurrent state updates to phase rotations on a complex unit circle, PMNet preserves gradient norms and inherently prevents divergence without the need for specialized initialization. We empirically demonstrate the active actuation of the memory module through a synthetic Copy-Paste task, where PMNet utilizes an expansive \textit{85-slot hierarchical memory tree} ($=\sum^{4}_{h=1}4^{h-1}$) to achieve near 100% exact retrieval across temporal distances that completely exceed the local sliding window attention's receptive field. Furthermore, despite being a compact 119M parameter model trained on 18.8B tokens, PMNet matches the zero-shot long-context robustness of a Mamba model that is three times larger. Our ablation studies and gradient analyses confirm that the historical failure of explicit memory was a structural alignment problem, which PMNet effectively overcomes, providing a theoretically grounded foundation for scalable sequence modeling.

Parameter-Efficient Fine-Tuning (PEFT), particularly Low-Rank Adaptation (LoRA), has become a standard approach for adapting Large Language Models (LLMs) under limited compute. However, in continual settings where models are updated sequentially with small datasets, conventional LoRA updates struggle to balance rapid adaptation and knowledge retention. Existing methods typically treat the low-rank space as a homogeneous update region, lacking mechanisms to regulate how short-term updates are consolidated over time. We propose a continual LoRA framework with \textbf{Pro}gram memory, inspired by \textbf{C}omplementary \textbf{L}earning Systems in neuroscience. Our approach, dubbed \textbf{ProCL}, organizes LoRA adapters into structured program memory slots that are dynamically retrieved through input-conditioned attention. This enables rapid and localized adaptation, encouraging similar inputs to reuse shared adapter regions while reserving unused capacity for future data. The slots are then combined with the underlying adapter, which maintains a distributed representation that gradually accumulates knowledge across tasks to balance plasticity and stability. Our method operates entirely within the LoRA parameterization and incurs no additional inference cost. Experiments on diverse benchmarks demonstrate improved retention and reduced catastrophic forgetting over other continual LoRA strategies.

Large language models increasingly need to accumulate and reuse historical information in long-term assistants and agent systems. Simply expanding the context window is costly and often fails to ensure effective context utilization. We propose $δ$-mem, a lightweight memory mechanism that augments a frozen full-attention backbone with a compact online state of associative memory. $δ$-mem compresses past information into a fixed-size state matrix updated by delta-rule learning, and uses its readout to generate low-rank corrections to the backbone's attention computation during generation. With only an $8\times8$ online memory state, $δ$-mem improves the average score to $1.10\times$ that of the frozen backbone and $1.15\times$ that of the strongest non-$δ$-mem memory baseline. It achieves larger gains on memory-heavy benchmarks, reaching $1.31\times$ on MemoryAgentBench and $1.20\times$ on LoCoMo, while largely preserving general capabilities. These results show that effective memory can be realized through a compact online state directly coupled with attention computation, without full fine-tuning, backbone replacement, or explicit context extension.

Robust optimization (RO) provides a principled framework for decision-making under uncertainty, but its practical use is often limited by the need to manually reformulate uncertain optimization models into tractable deterministic counterparts. Recent large language models (LLMs) have been shown promising for automating optimization formulation, yet RO reformulation remains challenging because it requires precise multi-step reasoning and mathematically consistent transformations. To facilitate systematic evaluation of LLM-based reformulation, for which no dedicated benchmark currently exists, we develop AutoRO-Bench, a benchmark featuring an automated data generation pipeline for the core RO reformulation task and a curated dataset for the RO application task. To address the reformulation challenge, we propose Automated Reformulation with Experience Memory (AutoREM), a tuning-free memory-augmented framework that autonomously builds a structured textual experience memory by reflecting on past failed trajectories through a tailored offline adaptation procedure. AutoREM requires neither domain-specific expert knowledge nor parameter updates, and the resulting memory readily transfers across different base LLMs. Experimental results show that AutoREM consistently improves the accuracy and efficiency of RO reformulation across in-distribution datasets, out-of-distribution datasets, and diverse base LLMs.

Memory consolidation, the process by which transient experiences are transformed into stable, structured representations, is a foundational organizing principle in the human brain, yet it remains largely unexplored as a design principle for modern sequence models. In this work, we leverage established neuroscientific theories of memory consolidation and cross-frequency coupling to propose the Hierarchical Memory Module (HMM), a neural memory architecture composed of two functionally distinct sub-modules that operate at different update frequencies. Inspired by the transformation hypothesis, the low-frequency sub-module produces high-level representations that capture abstract, gist-level knowledge, while the high-frequency sub-module produces fine-grained representations that preserve richer episodic detail. The final memory output is dynamically reconstructed as a context-dependent combination of both representations, analogous to the reconstructive nature of human memory retrieval. We integrate HMM into a Transformer-based language decoder to form Mela, a family of memory-augmented language models that perform online memory consolidation at test time. To further exploit the multi-granularity memory representations produced by HMM, we introduce MemStack, a method that distributes different levels of memory features across the early layers of the decoder without introducing additional tokens. Experiments on language modeling demonstrate that Mela outperforms Transformer baselines across all the model sizes. Moreover, with the pretrained context length fixed at 4K, Mela maintains performance on significantly longer contexts, whereas Transformer baselines degrade rapidly beyond their training length. Extensive ablation studies validate the contribution of each component and provide guidance for practical configuration.

Text embedding and generative tasks are usually trained separately based on large language models (LLMs) nowadays. This causes a large amount of training cost and deployment effort. Context compression is also a challenging and pressing task, which is vital to reasoning-driven generation, and agentic tasks requiring long context and continual learning. In this paper, we explore how to unify reasoning-driven generation, reasoning-enhanced text representation and context compression tasks in one forward pass for LLMs. Through meta latent tokens and a unified generative, representative and compressive tuning approach, we propose a training framework named GRC that bridges the three tasks. The trained models can accomplish three objectives in a single forward pass while maintaining modular, LEGO-style flexibility during inference. This design greatly reduces the deployment effort for retrieval-augmented generation (RAG) and achieves efficient inference and three times data utilization during training. Furthermore, this framework design enables a new paradigm for text embedding: self-reason-latent embeds, and a new generation paradigm, latent memory-augmented generation, where compressed and internalized KV cache with O(1) length is used as the updatable memory. We also propose hybrid paged attention to speed up the inference of our models. Extensive experiments on reasoning-intensive retrieval benchmarks, generative tasks, document compression, latency evaluation, and RAG settings demonstrate the effectiveness of our method and may shed light on the truly unified model that can handle reasoning-driven generation, embedding and compression tasks seamlessly.

Standard Transformers impose near-exponential decay on the influence of distant tokens, conflicting with the power-law structure of long-range dependencies in natural language. We introduce the \emph{Variable-Order Retention Transformer} (\VORT{}), a memory architecture in which each ingested token is assigned a learnable fractional order α_i\in[δ,1] that governs a Grünwald--Letnikov power-law retention kernel. Because the fractional weighted sum is non-Markovian, we approximate it through a sum-of-exponentials (SOE) decomposition computed by Gauss--Laguerre quadrature on a Laplace-type integral representation of the kernel weights. Each exponential component admits a one-step Markovian recurrence at O(Sd_v) per step, where S=O(\log(T/\varepsilon)) terms suffice for \varepsilon-uniform accuracy on horizon [1,T]. Retrieval is keyed and associative via a linear-attention accumulator with an exact O(KSd_φd_v) -per-step recurrence. Four results are established: (i) an SOE approximation theorem with geometric convergence rate from the analyticity of the integrand after a log-change of variables; (ii) a quantisation bound valid on [δ,1] with correct analysis near α=0; (iii) a direct L^2 energy argument (Proposition) showing that for α>1/2 any mixture with fixed minimum decay rate Λ>0 incurs L^2([1,T]) error at least N_α(T)-C(Λ)\to\infty, with the Λ-dependence made explicit; and (iv) linear convergence of a gradient plasticity rule under the Polyak--Łojasiewicz condition. Two synthetic experiments confirm the architectural advantage: a Zipf-distributed retrieval benchmark and an entity label-copy task with uniform lag distribution, the latter ruling out prior-matching as an explanation for the power-law kernel's advantage.

Current language model memory systems store what happened but not how it felt. This distinction -- between semantic memory (knowing about a past event) and episodic memory (re-experiencing it) -- was identified by Tulving as the difference between noetic and autonoetic consciousness. Damasio demonstrated that humans with intact knowledge but absent emotional markers exhibit impaired decision-making. We bridge this gap for language models. Using Gemma 3 1B-IT with pretrained Gemma Scope 2 sparse autoencoders, we identify 310 emotion-exclusive features at layer 22 with psychologically valid geometry. We construct distinctive-feature emotion vectors during experience and partially re-inject them during recall, triggered by context similarity at layer 7. We test four conditions paralleling Damasio's framework: A (no memory), B (semantic labels), C (emotion echo), and BC (semantic + echo). For emotional orientation, the echo alone steepens the threat-safety gradient: the regression slope of threat rating on contextual similarity is 0.80 for C vs 0.56 for A ($p$=0.011, permutation test). For decisions, the echo amplifies knowledge into action: BC=80% good choices vs B=52% ($z$=+2.60, $p$<0.01), while the echo alone has no effect (C=22%, n.s.). The echo changes how the model feels independently, but changes what it does only when combined with knowledge -- replicating Damasio's core finding. The echo amplifies knowledge. It does not replace it.

We present Grokers, an architecture for building persistent, structured comprehension of typed knowledge graphs through bottom-up inductive traversal of dependency subgraphs. Unlike retrieval-augmented generation (RAG), which pays full comprehension cost at every query, Grokers pushes intelligence to write time: autonomous Groker agents analyze nodes in a typed stream graph, extract structured attributes via governed language model (LM) calls, and inductively compose that understanding upward through dependency relations, writing enriched typed attributes that serve all future queries at zero additional LM cost. We prove three formal properties: (1) the Byte-Identity Theorem, establishing that context blocks assembled from a transactionally-maintained denormalization index are byte-identical across LM turns between semantic changes, enabling KV-cache hit rates approaching 100%; (2) the Accumulation Monotonicity Theorem, establishing that the fraction of interactions resolved without LM calls is non-decreasing in the number of completed interactions under a governed wisdom library growth protocol; and (3) the Dual-Traversal Ordering Theorem, establishing that top-down generation and bottom-up comprehension are the unique correct traversal orderings for their respective tasks over a dependency DAG, and that their composition closes into a complete generation-comprehension cycle. We further present a deterministic alternative to embedding-based semantic search, with a synonym caching protocol whose LM fallback rate converges to zero for finite-vocabulary domains. A reference implementation is provided in the open-source Qbix / Safebox / Safebots stack.

Steering large language models (LLMs) is usually done by either instruction prompting or activation steering. Prompting often gives strong control, but caches guidance tokens at every layer and can clutter long interactions; activation steering is compact but typically weaker and does not support large structured reminders. We introduce memory inception (MI), a training-free method that steers in latent attention space by inserting text-derived key-value (KV) banks only at selected layers. Rather than materializing reminder content throughout the prompt cache, MI treats steering as selective KV allocation, injecting latent slots only where the model routes to them. On matched personality-steering tasks, MI gives the best overall control--drift trade-off, remaining competitive with prompting while consistently outperforming CAA. On updateable guidance, MI supports mid-conversation behavior shifts without rewriting the visible transcript, achieving the highest post-shift alignment on Qwen3. On structured reasoning, MI outperforms visible prompting on HARDMath and PHYSICS (10/12 subject$\times$mode cells), serving as proxies for structured reasoning in verifiable domains, while cutting content-matched KV storage by up to 118$\times$. These results position MI as a powerful steering method when guidance is persistent, structured, or expensive to keep in the visible transcript.

We revisit a universally accepted but under-examined design choice in every modern LLM: a token index is looked up once at the input embedding layer and then permanently discarded. This single-injection assumption induces two structural failures: (i) the Rare Token Problem, where a Zipf-type distribution of vocabulary causes rare-token embeddings are chronically under-trained due to receiving a fraction of the cumulative gradient signal compared to common tokens; and (ii) the Contextual Collapse Problem, where limited parameters models map distributionally similar tokens to indistinguishable hidden states. As an attempt to address both, we propose TIDE, which augments the standard transformer with EmbeddingMemory: an ensemble of K independent MemoryBlocks that map token indices to context-free semantic vectors, computed once and injected into every layer through a depth-conditioned softmax router with a learnable null bank. We theoretically and empirically establish the benefits of TIDE in addressing the issues associated with single-token identity injection as well as improve performance across multiple language modeling and downstream tasks.

While Separate Source-Channel Coding (SSCC) retains the practical benefits of modular system design, its effectiveness in noisy text transmission is fundamentally constrained by the fragility of autoregressive source decoding. In low-SNR regimes, even a small number of residual bit errors after channel decoding may derail the subsequent lossless reconstruction process, especially when Arithmetic Coding (AC) relies on Large Language Model (LLM)-based probability estimation. Existing remedies either strengthen channel decoding based solely on channel observations or introduce contextual information only at the receiver for post-hoc correction, yet neither fully addresses the fragility of source probability modeling under residual channel errors. To this end, this paper proposes a Memory-Augmented Source Coding (MASC) scheme for robust SSCC-based transmission. Rather than treating context as external side information, MASC internalizes contextual patterns into a source model shared by both the transmitter-side source encoder and the receiver-side source decoder. Specifically, MASC employs a shared Parameterized Contextual Memory (PCM) to encode multi-order $n$-gram patterns, and further introduces a Mixture-of-Memory-Experts Router (MMER) to perform sparse, hidden-state-dependent routing over memory experts during autoregressive source modeling. By adaptively activating only the most relevant memories at each coding step, MASC refines source probability estimation, shortens average codelength, and mitigates the sensitivity of source decoding to residual channel errors. Extensive experiments over Rayleigh fading and AWGN channels demonstrate the effectiveness of the proposed scheme compared with state-of-the-art methods.

Adapting a pretrained language model to a new task often hurts the general capabilities it already had, a problem known as catastrophic forgetting. Sparse Memory Finetuning (SMF) tries to avoid this by adding key-value memory layers to the model and, on each training step, updating only the small set of memory rows that the current batch reads most heavily. We re-implement SMF on Qwen-2.5-0.5B-Instruct and compare it with LoRA and full finetuning on MedMCQA, a 4-choice medical exam task, using WikiText perplexity and TriviaQA accuracy as forgetting probes. SMF improves MedMCQA by 2.5 percentage points while keeping both forgetting probes within roughly 1 point of the base model, whereas LoRA and full finetuning achieve larger gains but with clear drift on both. We also compare two row-selection rules (KL-divergence and TF-IDF), which balance the two forgetting metrics differently.

Token communication has emerged as a promising framework for efficient wireless transmission by representing source data as compact semantic tokens. However, transmitting full semantic tokens still incurs considerable communication overhead. In this paper, we propose an evolving semantic token communication system with a parametric memory network over MIMO fading channels. Specifically, only an equal-length prefix of each semantic token is transmitted, which reduces transmission cost while preserving a consistent token structure for receiver-side recovery. At the receiver, a parametric memory network is introduced to reconstruct the missing suffix information from the received token prefixes, where semantic memory is stored implicitly in the network parameters. To realize this design, full semantic tokens are first organized into a codebook, and truncated tokens are paired with the codeword labels of their corresponding full tokens. Based on these token-label pairs, kNN-based teacher distributions are constructed to fine-tune a pretrained GPT-2-based recovery module, which learns to infer the codeword distribution of each incomplete token and recover the corresponding complete semantic token. In addition, an online evolution strategy is developed to periodically update the parametric memory network and the entire system using newly observed test samples, thereby improving adaptability under distribution shifts. Experimental results demonstrate that the proposed method consistently outperforms the existing evolving memory benchmark under different channel conditions and channel bandwidth ratios, with up to 1.09 dB PSNR improvement.

Large language models encode vast factual knowledge that can become outdated or incorrect after deployment, yet retraining is prohibitively costly. This motivates lifelong model editing, which updates targeted behavior while preserving the rest of the model. Existing editors, both parameter-modifying and parameter-preserving, degrade severely as edits accumulate and struggle to generalize across paraphrases. We propose HoReN, a codebook-based parameter-preserving editor that wraps a single MLP layer with a discrete key-value memory. HoReN treats each codebook entry as both a knowledge key and a Hopfield stored pattern, retrieves edits by angular similarity on the unit hypersphere, and refines queries through damped Hopfield dynamics so paraphrases converge to the correct memory basin while unrelated inputs remain stable. HoReN achieves strong editing performance with consistent gains across diverse benchmarks spanning standard ZsRE, structured WikiBigEdit, and unstructured UnKE evaluations. Moreover, HoReN scales to 50K sequential edits on ZsRE with stable overall performance above 0.93, while prior editors collapse or degrade severely before reaching 10K. Our code is available at https://github.com/ha11ucin8/HoReN.

High-precision medical diagnosis relies not only on static imaging features but also on the implicit diagnostic memory experts instantly invoke during image interpretation. We pinpoint a fundamental cognitive misalignment in medical VLMs caused by discrete tokenization, leading to quantization loss, long-range information dissipation, and missing case-adaptive expertise. To bridge this gap, we propose ours, a framework for latent diagnostic memory evolution that simulates the experiential invocation of clinicians by dynamically synthesizing implicit diagnostic memories within the model's hidden stream. Specifically, it begins with a Meta Query for Prior Memorization mechanism, where learnable probes retrieve structured priors from an anatomical prior encoder to generate condensed implicit memories. To ensure clinical fidelity, we introduce Causal Counterfactual Refinement (CCR), which leverages reinforcement learning and counterfactual rewards derived from region-level feature masking to quantify the causal contribution of each memory, thereby pruning redundancies and aligning latent representations with diagnostic logic. This evolutionary process culminates in Intrinsic Memory Transition (IMT), a privileged-autonomous dual-branch paradigm that internalizes teacher-branch diagnostic patterns into the student-branch via full-vocabulary divergence alignment. Comprehensive empirical evaluations across multiple datasets demonstrate that ours, by transferring external expertise into endogenous parameters, significantly outperforms existing state-of-the-art methods, particularly chain-of-thought paradigms, in diagnostic accuracy. The code is available at https://github.com/zhcz328/MedSynapse-V.

Large Language Models (LLMs) have achieved strong performance on general code generation, but their effectiveness drops sharply in enterprise settings where software development relies on internal private libraries absent from public pre-training corpora. Existing Retrieval-Augmented Generation (RAG) methods provide a training-free solution by retrieving static API documentation, but our analysis shows that documentation mainly helps models identify what APIs to use and remains insufficient for teaching how to use them correctly. Even with oracle API-document retrieval, LLMs still make recurring errors at the API, cross-API, and task levels, including API misuse or hallucination, flawed API composition, and incorrect solution strategies. To address this limitation, we propose MEMCoder, a training-free self-evolving memory framework for private-library code generation. MEMCoder augments existing RAG pipelines with a Multi-level Evolving Memory that continuously accumulates and reuses execution-derived Usage Guidelines at the API, cross-API, and task levels. During generation, MEMCoder retrieves both static API documentation and relevant historical memories to guide code generation; after execution, it analyzes feedback to refine memory through a closed loop of generation, execution, reflection, and update. Extensive experiments on NdonnxEval and NumbaEval show that MEMCoder consistently enhances different RAG backbones across LLMs of different scales, yielding an average absolute pass@1 improvement of 18.41 percentage points. Moreover, MEMCoder outperforms existing self-evolving memory methods and validates the effectiveness of organizing execution feedback into multi-level usage memories.

We investigate whether the Feed-Forward Network (FFN) sublayer in a decoder-only transformer can be replaced by an explicit learned memory graph while preserving the surrounding autoregressive architecture. The proposed Graph Memory Transformer (GMT) keeps causal self-attention intact, but replaces the usual per-token FFN transformation with a memory cell that routes token representations over a learned bank of centroids connected by a learned directed transition matrix. In the base GMT v7 instantiation studied here, each of 16 transformer blocks contains 128 centroids, a 128 * 128 edge matrix, gravitational source routing, token-conditioned target selection, and a gated displacement readout. The cell therefore returns movement from an estimated source memory state toward a target memory state, rather than a retrieved value. The resulting model is a fully decoder-only language model with 82.2M trainable parameters and no dense FFN sublayers, compared with a 103.0M-parameter dense GPT-style baseline used in the evaluation. The base v7 model trains stably and exposes centroid usage, transition structure, and source-to-target movement as directly inspectable quantities of the forward computation. It remains behind the larger dense baseline in validation loss and perplexity (3.5995/36.58 vs. 3.2903/26.85), while showing close zero-shot benchmark behavior under the evaluated setting. These results are not intended as a state-of-the-art claim; they support the viability and structural interpretability of replacing dense within-token transformation with graph-mediated memory navigation. Broader scaling, optimized kernels, and more extensive benchmark evaluation are left for subsequent work.

Evolve pairs a small local language model with a persistent, teacher-compiled knowledge store -- refined through sleep consolidation and usage-driven refresh -- to deliver substantial accuracy gains over the model's parametric baseline while amortizing teacher costs through cross-query knowledge reuse. Rather than retrieving document fragments at query time, Evolve constructs a store of semantically coherent sections compiled by teacher models at natural conceptual boundaries; new sections are staged on acquisition, consolidated offline through teacher-mediated merging, and refreshed inline when expired. A 2B-parameter local model handles classification and generation; large teacher models are invoked only for knowledge operations. Across 750 benchmark queries spanning custom specialist questions, NaturalQuestions, and TriviaQA, the 2B model augmented by Evolve improves from 20-33% baseline accuracy to 60-84% (+40-52pp) while reducing teacher invocations by over 50% through reuse. Post-consolidation compresses the knowledge store by 31-33.5% across three independent benchmarks while preserving accuracy; section-based retrieval outperforms chunk-based retrieval by 5-9pp across every lifecycle condition. The architecture supports two generation modes over the same lifecycle -- suppress (strict section-only grounding, auditable) and augment (section-supplemented responses).

Large token-indexed lookup tables provide a compute-decoupled scaling path, but their practical gains are often limited by poor parameter efficiency and rapid memory growth. We attribute these limitations to Zipfian under-training of the long tail, heterogeneous demand across layers, and "slot collapse" that produces redundant embeddings. To address this, we propose X-GRAM, a frequency-aware dynamic token-injection framework. X-GRAM employs hybrid hashing and alias mixing to compress the tail while preserving head capacity, and refines retrieved vectors via normalized SwiGLU ShortConv to extract diverse local n-gram features. These signals are integrated into attention value streams and inter-layer residuals using depth-aware gating, effectively aligning static memory with dynamic context. This design introduces a memory-centric scaling axis that decouples model capacity from FLOPs. Extensive evaluations at the 0.73B and 1.15B scales show that X-GRAM improves average accuracy by as much as 4.4 points over the vanilla backbone and 3.2 points over strong retrieval baselines, while using substantially smaller tables in the 50% configuration. Overall, by decoupling capacity from compute through efficient memory management, X-GRAM offers a scalable and practical paradigm for future memory-augmented architectures. Code aviliable in https://github.com/Longyichen/X-gram.

This paper presents a hybrid architecture for intelligent systems in which large language models (LLMs) are extended with an external ontological memory layer. Instead of relying solely on parametric knowledge and vector-based retrieval (RAG), the proposed approach constructs and maintains a structured knowledge graph using RDF/OWL representations, enabling persistent, verifiable, and semantically grounded reasoning. The core contribution is an automated pipeline for ontology construction from heterogeneous data sources, including documents, APIs, and dialogue logs. The system performs entity recognition, relation extraction, normalization, and triple generation, followed by validation using SHACL and OWL constraints, and continuous graph updates. During inference, LLMs operate over a combined context that integrates vector-based retrieval with graph-based reasoning and external tool interaction. Experimental observations on planning tasks, including the Tower of Hanoi benchmark, indicate that ontology augmentation improves performance in multi-step reasoning scenarios compared to baseline LLM systems. In addition, the ontology layer enables formal validation of generated outputs, transforming the system into a generation-verification-correction pipeline. The proposed architecture addresses key limitations of current LLM-based systems, including lack of long-term memory, weak structural understanding, and limited reasoning capabilities. It provides a foundation for building agent-based systems, robotics applications, and enterprise AI solutions that require persistent knowledge, explainability, and reliable decision-making.

We present SCM (Sleep-Consolidated Memory), a research preview of a memory architecture for large language models that draws on neuroscientific principles to address a fundamental limitation in current systems: the absence of persistent, structured, and biologically plausible memory. Existing approaches rely on truncating context windows, growing vector databases without bound, or tiered storage systems that lack consolidation and forgetting mechanisms. SCM implements five core components inspired by human memory: a limited-capacity working memory, multi-dimensional importance tagging, offline sleep-stage consolidation with distinct NREM and REM phases, intentional value-based forgetting, and a computational self-model enabling introspection. Across a standardized benchmark suite of eight tests, the prototype achieves perfect recall accuracy over ten-turn conversations while reducing memory noise by 90.9% through adaptive forgetting. Memory search latency remains below one millisecond even with hundreds of stored concepts. This work establishes the architectural foundations for memory systems that consolidate, prioritize, and forget, offering a testable platform for advancing LLM memory research.

Large language models (LLMs) encode knowledge in parametric weights, making it costly to update or extend without retraining. Retrieval-augmented generation (RAG) mitigates this limitation by appending retrieved text to the input, but operates purely through context expansion, where external knowledge competes as tokens within the attention mechanism. As a result, its influence is indirect and often unstable, particularly in long context and multi hop reasoning scenarios. We propose Knowledge Capsules, structured nonparametric memory units that represent normalized relational knowledge and can be constructed directly from document corpora using a frozen base model. Instead of injecting knowledge as text, we introduce an External Key Value Injection (KVI) framework that compiles capsules into attention-compatible key value representations, enabling external knowledge to directly participate in the model's attention computation. By shifting knowledge integration from context-level augmentation to memory level interaction, the proposed framework consistently outperforms RAG and GraphRAG across multiple QA benchmarks, with improved stability and accuracy in long context and multi hop reasoning, while requiring no parameter updates.

Transformers suffer from a high computational cost that grows with sequence length for self-attention, making inference in long streams prohibited by memory consumption. Constant-memory alternatives such as RNNs and SSMs compress history into states with fixed size and thus lose long-tail dependencies, while methods that memorize contexts into parameters, such as Test-Time Training (TTT), are prone to overfitting token-level projection and fail to preserve the causal effect of context in pretrained LLMs. We propose Absorber LLM, which formulates long-context retention as a self-supervised causal synchronization: after absorbing historical contexts into parameters, a contextless model should match the original model with full context on future generations. We optimize this objective by synchronizing internal behaviors of the updated model with the original one, ensuring context absorption and generalization. Experiments on long-context and streaming benchmarks show that Absorber LLM reduces inference memory and improves accuracy over prior parameter-as-memory baselines.

Prompt-injected memory can improve reasoning without updating model weights, but it also creates a control problem: retrieved content helps only when it is applied in the right state. We study this problem in a strict training-free setting and formulate it as applicability control: when to trigger a memory-assisted second pass, when to trust it, and how to maintain the memory bank over time. Our method combines uncertainty-based routing, confidence-based selective acceptance, bank selection across rule and exemplar memory, and evidence-based governance of the memory bank over time. Under a locked training-free protocol with compute-matched controls, it improves two core arithmetic benchmarks by +7.0 points on SVAMP and +7.67 points on ASDiv over baseline. The same architecture also transfers to QA and agent benchmarks with smaller positive effects and shows the same positive direction on a second checkpoint for the main arithmetic tasks. On arithmetic, the main empirical pattern is that the control architecture, rather than raw memory exposure, drives the improvements on SVAMP and ASDiv. Mechanistically, confidence separates helpful from harmful rule-bank interventions, and under fixed retrieval the repair-versus-corrupt difference localizes to rows whose retrieved set actually contains the edited entries.

While large language models have achieved remarkable performance in complex tasks, they still need a memory system to utilize historical experience in long-term interactions. Existing memory methods (e.g., A-Mem, Mem0) place excessive emphasis on organizing interactions by frequently rewriting them, however, this heavy reliance on summarization risks diluting essential contextual nuances and obscuring key retrieval features. To bridge this gap, we introduce AnchorMem, a novel memory framework inspired by the Proust Phenomenon in cognitive science, where a specific anchor triggers a holistic recollection. We propose a method that decouples the retrieval unit from the generation context. AnchorMem extracts atomic facts from interaction history to serve as retrieval anchors, while preserving the original context as the immutable context. To reveal implicit narrative cues, we construct an associative event graph that uses higher-order event links that bind sets of related facts into shared event representations, strengthening cross-memory integration without relying on generic entities as bridges. During retrieval, the system anchors queries to specific facts and events to locate relevant memories, but reconstructs the context using the associated raw chunks and events. Our method reconciles fine-grained retrieval with the contextual integrity of interactions. Experiments across three closed-source and open-source models on the LoCoMo benchmark demonstrate that AnchorMem significantly outperforms baselines. Code is available at https://github.com/RayNeo-AI-2025/AnchorMem.

The most important architectural problem in AI is not the size of the model but the absence of a layer that carries forward what the model has come to understand. Sessions end. Context windows fill. Memory APIs return flat facts that the model has to reinterpret from scratch on every read. The result is intelligence that is powerful per session and amnesiac across time. This position paper argues that the layer which fixes this, the continuity layer, is the most consequential piece of infrastructure the field has not yet built, and that the engineering work to build it has begun in public. The formal evaluation framework for the property described here is the ATANT benchmark (arXiv:2604.06710), published separately with evaluation results on a 250-story corpus; a companion paper (arXiv:2604.10981) positions this framework against existing memory, long-context, and agentic-memory benchmarks. The paper defines continuity as a system property with seven required characteristics, distinct from memory and from retrieval; describes a storage primitive (Decomposed Trace Convergence Memory) whose write-time decomposition and read-time reconstruction produce that property; maps the engineering architecture to the theological pattern of kenosis and the symbolic pattern of Alpha and Omega, and argues this mapping is structural rather than metaphorical; proposes a four-layer development arc from external SDK to hardware node to long-horizon human infrastructure; examines why the physics limits now constraining the model layer make the continuity layer newly consequential; and argues that the governance architecture (privacy implemented as physics rather than policy, founder-controlled class shares on non-negotiable architectural commitments) is inseparable from the product itself.