Overview

Continual Learning studies how models adapt to sequences of tasks or non-stationary data without losing useful earlier knowledge, a challenge known as catastrophic forgetting. Common approaches replay selected examples, regularize updates to protect important parameters, isolate tasks with adapters or separate subnetworks, or manage knowledge through prompts, memory, and model expansion. Methods are tested on class- and domain-incremental image benchmarks such as CIFAR and ImageNet, as well as task-free streams and applications in language, robotics, and healthcare. A central tension is balancing retention with the capacity to learn, while evaluations increasingly question assumptions such as known task boundaries and seek to measure transfer, recovery, resource use, and performance under realistic shifts.

Latest papers

323 papers26 papers in the last four weeks, down 24% on the four weeks before. 0.2% of all new papers.

Oct 13, 2026

Stable-Plastic Subspace Construction for Differentially Private Continual LoRA

Continual learning (CL) adapts pretrained models to a sequence of tasks while retaining knowledge acquired from earlier tasks. When task data are sensitive, differential privacy (DP) provides record-level protection, but maintaining both adaptation and retention under privacy constraints remains challenging. We study this problem in continual low-rank adaptation (LoRA) and observe that: 1) private training increases forgetting primarily during subsequent adaptation rather than initial task acquisition; 2) jointly perturbing both LoRA factors introduces a non-negligible bilinear noise interaction; and 3) DP aggravates the stability-plasticity dilemma, hindering both knowledge retention and adaptation to new tasks. Based on these observations, we propose DP-SPA, which constructs a stable subspace from previously privatized updates and a plastic subspace from pretrained weights within its orthogonal complement. The method fixes these bases and privately optimizes only the LoRA factor, eliminating the bilinear noise interaction within each step. Before merging each task update, it rescales stable-direction coefficients using historical and current update energies. We establish a privacy guarantee and show that recalibration does not increase historical overlap and tightens a conditional upper bound on old-task loss increase. Experiments across five datasets, eight settings, and multiple privacy budgets show that DP-SPA improves average accuracy and reduces forgetting compared with the evaluated baselines.
Oct 12, 2026

Effect of Spherical Parameter Space on Number of Non-Forgetting Tasks and Capacity of Continual Learning

We propose a continuous learning formulation for TT non-forgetting tasks with the spherical parameter space Sd−1(d)\mathbb{S}^{d-1}(\sqrt{d}) in asymptotic regime d→∞d\rightarrow \infty. Each task ii has an optimum Θi\bmΘ_i, minimizing its corresponding loss function Ri(w)R_i(\bm{w}) on sphere. We reduce each task ii to an acceptable loss region modeled by a geodesic ball BΘi(ri(d))B_{\bmΘ_i}(r_i(d)) on Sd−1(d)\mathbb{S}^{d-1}(\sqrt{d}) centered at Θi\bmΘ_i with radius ri(d)r_i(d). To model task dependencies (similarities), task optima Θi\bmΘ_i are distributed by a Neyman-Scott clustered point process ΦΦ. Cluster centers {Yj}j=1Np\{\bm{Y}_j\}_{j=1}^{N_p} form a homogeneous Poisson point process on sphere, while intra-cluster task optima follow a von Mises-Fisher distribution with mean Yj\bm{Y}_j and concentration κκ. Each task t∈{1,...,T}t\in \{1, ..., T\} learning algorithm is considered to be a channel Qt(dwt∣wt−1,Kt−1(θ1:t−1),θt)Q_{t}(\mathrm{d}\bm{w}_t | \bm{w}_{t-1}, \mathcal{K}_{t-1}(\bmθ_{1:t-1}), \bmθ_t), where Wt\bm{W}_t is the learned parameter of task tt and Kt−1(Θ1:t−1)=⋂i=1t−1BΘi(ri(d))\mathcal{K}_{t-1}(\bmΘ_{1:t-1})=\bigcap_{i=1}^{t-1}B_{\bmΘ_i}(r_i(d)). Since in continual learning, the learner does not need to reach Θt\bmΘ_t exactly to successfully learn task tt, to avoid forgetting as well as reliable learning, it only needs for the learned parameter vector Wt\bm{W}_{t} to enter BΘt(rt(d))∩Kt−1(Θ1:t−1)B_{\bmΘ_t}(r_t(d)) \cap \mathcal{K}_{t-1}(\bmΘ_{1:t-1}). Using this model, we derive an upper bound on TT ensuring P[KT(Θ1:T)≠∅]→1\mathbb{P}[\mathcal{K}_{T}(\bmΘ_{1:T}) \neq \emptyset] \rightarrow 1 as d→∞d\rightarrow \infty and κ/d→ακ/d \rightarrow α, where α>0α>0 is a constant. Finally, for TT tasks sampled uniformly from ΦΦ, we derive an upper bound for total information theoretic capacity under the non-forgetting constraint for each task.
Oct 10, 2026

Engram Neuron Inspired Adaptive Memorization for Lifelong Learning

Lifelong learning enables artificial systems to adapt to changing environments. To achieve this, an effective artificial learner should dynamically reorganize past and new experiences, maintaining stability while enabling flexible knowledge integration. However, existing efforts primarily mitigate catastrophic forgetting, lacking the ability to adaptively update memory akin to biological systems. Here, by integrating the full memory life cycle of engram neurons into artificial learners, we propose a framework that constructs a structured memory from gradient representations, which regulates parameter updates and guides memory management. Through comprehensive evaluations on four benchmarks, our approach shows promising lifelong learning capability, maintaining stability over long task sequences and supporting flexible adaptation. Beyond the performance gains, our approach offers a computational perspective on robust and flexible lifelong learning, echoing mechanisms observed in biological memory systems.
Oct 8, 2026

LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs

Lifelong editing of LLMs requires storing thousands of edits after acquisition. A widely used family of approaches attaches one LoRA adapter per edit, which preserves behavior but grows linearly in storage. To address this challenge, we propose LadderEdit, a method that compresses each LoRA adapter after it is acquired. Each edit is first stored at low rank as a cheap sketch. We then check whether this sketch still satisfies the rewrite, generalization, and locality contract on probe prompts. Edits that pass keep the sketch; those that fail are promoted to a higher rank along a ladder until the contract is met. Because every edit retains some representation, coverage is maintained, and only hard edits consume more rank. Across ZsRE, CounterFact, and WikiBigEdit benchmarks on LLaMA-3-8B, Mistral-7B, and Qwen2.5-7B, LadderEdit tracks exact LoRA storage at 5.2x less memory and remains effective at 50,000 sequential edits.
Oct 7, 2026

Continual Learning without Continual Training

Continual learning requires models to adapt to new domains and new classes while retaining prior knowledge. Many existing methods rely on continued optimization, using regularization, replay, or parameter expansion to prevent new updates from overwriting previously learned knowledge. Instead, we propose replacing continual training with continual inference: a PFN-based model that is meta-trained, and then frozen, adapting to new classes only by extending an in-context evidence set. Our model, Latent Concept PFN, performs in-context Bayesian inference over a latent concept space that captures semantic structure shared across domains and classes. As each new domain or class arrives, exemplars are added to the memory; adaptation reflects updated posterior beliefs over latent concepts rather than gradient updates. No parameters are changed, reducing forgetting. The same method handles both domain and class incremental continual learning without task identity. Concept annotations are only used during meta-training, acting as a soft anchor on the latent space rather than a fixed bottleneck. Unlike fixed-vocabulary concept methods, the model also handles noisy, ambiguous, or incomplete annotations by combining concept labels with raw input evidence to discover distinctions beyond the predefined concept set. Experiments on class and domain incremental learning datasets demonstrate competitive continual learning performance while learning interpretable latent concepts.
Oct 4, 2026

MAGIC: Topology-Aware Analytic Graph Few-Shot Class-Incremental Learning

Graph few-shot class-incremental learning (GFSCIL) requires a model to continually recognize emerging classes from only a few labeled nodes while preserving previously acquired knowledge. Beyond the catastrophic forgetting inherited from conventional graph continual learning, GFSCIL presents two distinctive challenges: extremely limited novel-class supervision causes severe overfitting, while cross-session edges---edges connecting newly arriving nodes with historical nodes---alter historical propagation neighborhoods and thereby induce representation drift. We propose MAGIC, a replay-free GFSCIL framework that combines a frozen graph representation backbone (e.g., an intrinsically parameter-free backbone such as SGC or a pretrained graph foundation model) with closed-form analytic continual learning. To alleviate novel-session overfitting, MAGIC learns a topological prior from the base graph that can characterize both homophilous and heterophilous relations, and injects this prior through Potts Markov random field inference to refine supervision for novel classes. To mitigate representation drift, MAGIC transfers previous predictions from the old representations of affected historical nodes to their updated representations through drift-aware analytic distillation. Experiments across five datasets and eight baselines demonstrate the effectiveness of MAGIC. Under the 5-shot setting, MAGIC improves Mean Accuracy and Final Accuracy by 5.48 percentage points and 9.33 percentage points on average, and reduces Performance Drop by 10.78 percentage points on average compared with the best baselines. MAGIC also shows clear advantages under the 1- and 3-shot settings, with larger gains as the number of supports increases. Moreover, MAGIC requires substantially less training time.
Oct 3, 2026

Gated Target Propagation for Compositional Generalization in Continual Learning

Continual learning is typically framed as acquiring new knowledge without catastrophically forgetting previous tasks. However, a flexible continual learner should also be able to reuse and recombine previously acquired knowledge to rapidly solve novel task compositions. We introduce Gated Target Propagation (GaTaP), a continual learning algorithm in which task-specific gating variables---learned through a closed-form inner loop update---selectively suppress or enhance network modules. Network parameters are learned in a slower timescale outer loop, using the same local difference target propagation error signal as is used for adapting gating variables. We provide tractable experiments on class-incremental learning scenarios for both multilayer perceptron and convolutional network architectures. We show strong performance retention on previously learned tasks, as well as compositional generalization to unseen tasks, achieved through few-shot gain adaptation at inference. We analyze learned gating patterns and find that related tasks exhibit similar gating patterns, suggesting that inferred gates capture meaningful, reusable task structure. Overall, GaTaP provides a powerful framework for jointly ameliorating catastrophic forgetting and enabling few-shot compositional generalization in neural network models.
Oct 3, 2026

VoiceWeaver: Staged Learning of Structured Controls for Expressive Speech and Sound-Event Generation

Adding expressive and environmental controls to a speech generator requires learning heterogeneous attributes without losing earlier capabilities. VoiceWeaver addresses this problem with structured label prefixes and staged emotion-tone-event training in a shared text-audio model. Replay-based distillation retains earlier predictions, while embedding decorrelation and attribute dropout regularize conditioning. Evaluation separates single-attribute correctness, joint emotion-event generation, and three-attribute correctness. Rounded to whole percentage points, emotion accuracy is 86% in Chinese and 83% in English. Chinese emotion-event joint accuracy is about 70%, versus 56% for mixed-task training and 48% for Ming-omni-tts; English joint accuracy is 66%. Three-attribute accuracy is 59% in Chinese and 55% in English on 500 samples each (57% pooled). Text error rates exceed those of the external TTS baselines. Audio samples are available at https://anonymous.4open.science/api/repo/VoiceWeaver1-4D88/file/index.html.
Oct 3, 2026

PaLoRA: Paced Low-Rank Adaptation for Continual Learning

LoRA-based continual learning methods mitigate catastrophic forgetting through various mechanisms, yet nearly all complement these with small learning rates as a heuristic to restrict gradient scaling magnitude. Such fixed heuristics lack theoretical guidance on how the strength of this restriction should evolve as tasks accumulate. We reveal that even under directional constraints such as nullspace projection, finite-precision updates inevitably leak into the subspace of accumulated prior knowledge along multiple directions. While small learning rates attenuate such leakage, they cannot prevent the accumulated forgetting from intensifying as the effective rank of historical knowledge grows. We show that the optimal magnitude restriction should adaptively increase with this effective rank to balance stability and plasticity, i.e., preservation of previous knowledge and acquisition of new task information. Under an anisotropic leakage model, we derive a pacing law s∗=R/cs^*=\sqrt{R/c} that characterizes the optimal scaling of gradient steps, i.e., the magnitude restriction itself, where RR is the effective rank of past updates. Based on this insight, we propose PaLoRA, which compresses historical knowledge via adaptive SVD truncation, projects gradients onto the nullspace of prior tasks, and applies rank-aware adaptive pacing. Experiments demonstrate consistent improvements over prior methods, with particularly strong performance in long-horizon settings, achieving substantial gains of 4% accuracy on challenging 50-task ImageNet-A and ImageNet-R benchmarks.
Oct 2, 2026

One-Step Curvature Probes Miss the Fitting Operator: Retained Capacity and Terminal Null-Space Correction for Continual Learning

A one-step curvature probe evaluates an initial direction, whereas continual learners are judged after reaching comparable new-task fit. In an overparameterized linearization, projected gradient descent converges to ΔP=PJ⊤(JPJ⊤)−1rΔ_P=PJ^\top(JPJ^\top)^{-1}r, and its squared-displacement inflation is exactly the reciprocal of the retained fitting capacity cP(r)c_P(r). More generally, the terminal old-task quadratic ratio factorizes as 1/[cP(r)Gend(P,r)]1/[c_P(r)G_{\rm end}(P,r)], where GendG_{\rm end} compares curvature along endpoint directions. In the rank-one case, GendG_{\rm end} equals the one-step probe gain; for multiple outputs, the two gains can differ. The terminal quadratic also separates into a curvature-optimal fitting floor and an algorithm-dependent null-space excess, motivating terminal null-space correction, which preserves linearized new-task outputs on its Jacobian batch. Controlled checks validate the local quadratic and show that projection can greatly reduce matched-norm curvature while barely changing terminal forgetting. Across common-threshold configurations, projection yields the larger signed old-task loss change in 73/99 matched pairs, retained fitting capacity falls with rank, and fixed-rank comparisons separate terminal forgetting even when probe gain is approximately matched. On Permuted MNIST and Split CIFAR-100, terminal correction decreases signed old-task loss in 166/180 method--dataset--seed pairs under the all-seed intention-to-correct analysis; because acceptance and outcome reporting use the same held-out split, this result is descriptive and test-conditioned. Overall, terminal cost depends jointly on retained fitting capacity, endpoint-direction curvature, and the null-space component selected by optimization.
Oct 2, 2026

What Should World Models Forget? Stratified Retention for Continual Adaptation

Continual learning treats any drop in performance on previously seen data as forgetting, a convention inherited from settings where the prediction target is stationary. World models violate this assumption: their target is the environment, which changes, so knowledge that was accurate when learned can later become false, and revising it is correct behavior rather than a defect. Non-stationary ground truth is well studied in the concept drift literature and in the temporal factuality of language models, but it has not been formulated for world models, which are distinctive in that they also encode structural knowledge, such as physical laws and object permanence, that must never be degraded. We argue that continual world models need retention stratified by invariance timescale, the time over which knowledge stays true: invariants should be refined but never degraded, while instance-level facts should be revised once the environment changes. We show that standard forgetting metrics, used alone, count correct revision as forgetting and so rank a frozen model highest. We propose differential retention, which reports two quantities separately: invariant retention, checked after every adaptation round, and revision latency, the number of observations a model needs to pick up a change.
Oct 1, 2026

Task-Oriented Rank Adaptation for Continual Learning in Text Classification

Continual learning (CL) in text classification faces two critical challenges: catastrophic forgetting and negative transfer across sequential tasks. Parameter-Efficient Fine-Tuning (PEFT) methods such as LoRA enable efficient adaptation by learning low-rank updates of the model parameters. However, these compact representations are normally trained in isolation, limiting their reuse across related tasks. We introduce Task-Oriented Rank Adaptation (TORA), a geometric routing framework that leverages the low-rank structure of LoRA adapters to decide whether to transfer knowledge from the most compatible expert (Boosting) or isolate the new task (Shielding) based on structural similarity. Evaluated across 15 diverse text classification benchmarks, TORA consistently avoids harmful routing decisions: compatible tasks exceed their isolated performance while reducing training time, and structurally distant tasks are protected from interference with no loss in accuracy. With a single geometric threshold and no reliance on task identities or predefined sequences, TORA provides a simple and effective approach for dynamic adapter routing in sequential text classification systems.
Oct 1, 2026

MLoRA:Continual Low-Rank Adaptation via Relative-Response Subspace Selection and Momentum Orthogonalization

Low-rank adaptation (LoRA) provides a parameter-efficient approach to continual learning, but its nominal rank can conceal a loss of effective adaptation capacity. We identify spectral plasticity collapse: during sequential adaptation, update energy becomes concentrated in a small subset of singular modes, leaving much of the available low-rank space underutilized. This exposes a limitation of interference avoidance alone: protecting historical representations does not ensure that the remaining adaptation capacity is responsive to new tasks or effectively utilized. To address this problem, we propose MLoRA, which jointly controls capacity allocation and utilization. First, historical whitening identifies input directions with strong current-task response relative to accumulated historical response, yielding a task-adaptive basis that remains fixed during training. Second, approximate polar orthogonalization of momentum updates reduces spectral concentration within theselected space. An orthonormal basis connects these mechanisms by transferring the factor-update spectrum exactly tothe induced weight update. We establish a max--min characterization of exact subspace selection and derive cumulative spectral bounds under controlled cross-step anisotropy. Across five class-incremental benchmarks and eight incremental settings, MuLoRA achieves the highest mean accuracy in 15 of 16 reported metrics.
Oct 1, 2026

CLEAN: Psychometrically Consistent Incremental Cognitive Diagnosis under Concept-Space Expansion via Architectural Isolation

Cognitive diagnosis (CD) is a fundamental task in intelligent education that profiles learner proficiency over knowledge concepts. In real-world learning platforms, newly added items continually introduce previously unseen concepts, necessitating dynamic expansion of the underlying concept space. Yet existing incremental CD models assume a fixed concept space, allowing gradients from new items to overwrite historical pathways and induce catastrophic forgetting. More critically, these methods rely solely on soft constraints to preserve historical diagnoses. Such constraints may fail to satisfy the requirement of diagnostic invariance after incremental updates, a requirement known as psychometric consistency in cognitive diagnosis. Therefore, we propose CLEAN (Continual Learning with Expandable and Architecturally Isolated Networks), a novel incremental CD framework supporting concept-space expansion while providing structural guarantees for pointwise invariance of historical diagnoses. Specifically, CLEAN first introduces a strict topological bipartition protocol, freezes historical diagnostic functions and applies deterministic orthogonal column masking to sever gradient interference. Second, to accommodate concept expansion, expandable full-rank branches with micro-variance initialization are deployed to learn novel concepts. Finally, to verify that this architectural design achieves invariance by construction, we formalize Representation Drift (RD) to quantify the perturbation of historical traits. Extensive experiments on three large-scale educational datasets demonstrate that CLEAN achieves zero RD, preserving old-item metrics identically to static anchors through architectural isolation while remaining competitive with or superior to strong continual-learning baselines on new items.
Sep 30, 2026

Dynamic LoRA-Experts and Prototype-Ensemble Matching for Class-Incremental Learning

Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. Parameter-efficient fine-tuning with pre-trained models reduces parameter overhead but can suffer from cumulative interference and suboptimal alignment between inference samples and specialized modules. We propose Dynamic LoRA-Experts and Prototype-Ensemble Matching (DLEPEM), a two-stage rehearsal-free framework. DLEPEM allocates a task-specific LoRA-Expert for each incremental task to reduce cross-task interference, then combines frozen pre-trained-model prototypes with task-adaptive LoRA-Expert prototypes for reliable task-level discrimination. Experiments on standard CIL and Few-Shot CIL benchmarks demonstrate strong performance under the evaluated protocols.
Sep 30, 2026

SE-ADD: Self-Evolving Audio Deepfake Detection with Mistake-Driven Supervision

Audio deepfake detection (ADD) must remain effective when new spoofing attacks emerge after deployment. Emerging audio language model (ALM)-based ADD methods are built on predefined supervision from ground-truth labels or verified forensic rationales. However, this paradigm overlooks an ALM's own mistakes, which indicate where targeted supervision is most needed. To this end, we first introduce evolving spoofing environments for ALM-based ADD, where a new attack becomes dominant while previously observed attacks persist. Motivated by the above learning-from-mistakes perspective, we further propose SE-ADD, a self-evolving framework that iteratively adapts an ALM via low-rank adaptation (LoRA) using mistake-driven supervision built from its verdicts and self-generated forensic cues. All training samples receive direct authenticity supervision, while misclassified ones receive additional cue-augmented supervision. As verdicts and cues are regenerated by the updated ALM, the resulting supervision evolves accordingly. Experiments on two ALMs demonstrate the effectiveness of SE-ADD in generalizing to unseen attacks, reducing the equal error rate (EER) from 36.72%36.72\% to 7.52%7.52\% for Qwen2-Audio and from 19.93%19.93\% to 3.97%3.97\% for MOSS-Audio.
Sep 30, 2026

Hyperbolic Prototype Routing for Rehearsal-Free Class-Incremental Learning

Class-Incremental Learning (CIL) aims to continually learn new classes while preserving prior knowledge. Parameter-efficient fine-tuning with pre-trained models enables CIL with minimal parameter updates, but existing approaches still suffer from catastrophic forgetting caused by cumulative interference and suboptimal module-sample matching at inference. We propose Hyperbolic Prototype Routing (HyPro), a rehearsal-free framework for continual learning. HyPro allocates a dedicated LoRA-Expert module to each incremental task for isolated representation learning, then projects routing features onto a Poincare ball and performs geodesic nearest-prototype matching for reliable task-level discrimination. Extensive experiments on standard CIL and Few-Shot CIL benchmarks show that HyPro consistently improves average and final-stage accuracy over strong baselines.
Sep 30, 2026

A Dynamical Theory of LoRA in Continual Learning

Despite the widespread use of Low-Rank Adaptation (LoRA), little is known about its dynamics in continual learning and the mechanisms by which low-rank updates affect catastrophic forgetting. We provide an asymptotically exact dynamical characterization of LoRA in a solvable two-task teacher-student model. In the high-dimensional online-learning limit, we derive a closed system of ordinary differential equations for a finite set of macroscopic order parameters, yielding exact expressions for the generalization errors throughout both the initial Task 1 learning phase and the subsequent LoRA fine-tuning on Task 2. The theory quantitatively matches finite-dimensional simulations and exposes two characteristic effects of LoRA: low-rank adaptation reduces interference with features learned on the first task, but its initialization slows adaptation to the second task. Building on this mechanistic picture, we analyze a state-dependent masking strategy that freezes hidden units carrying the strongest first-task representations and restricts adaptation to the complementary subspace. This structural partitioning markedly reduces forgetting, while preserving plasticity on the new task. Our framework further clarifies the role of adapter rank: transfer improves only up to the intrinsic dimensionality of the target task and saturates beyond it, while forgetting continues to grow with rank. These results provide a dynamical and geometric account of how low-rank adaptation organizes information across sequential tasks and are qualitatively reproduced on a sequential MNIST benchmark.
Sep 30, 2026

Generalized Residual Closure: General Learning Dynamics for Stability-Plasticity Compatibility

Learning must acquire new capabilities while preserving both prior responsibilities and the capacity to learn again. We introduce Generalized Residual Closure (GRC), a framework for learning as recursive closure of future-relevant discrepancies: closing a residual establishes the conditions for subsequent prediction, interaction, and learning. Under a complete representation-relation description at a fixed learner-world boundary, persistent internal learning has two primitive modes: Transformation within a representation and revision of the Representation itself. We establish a local tangent decomposition under regularity assumptions and a criterion for when representation revision is necessary. In an affine model, we derive a necessary-and-sufficient condition for stability-plasticity compatibility and the unique solution of a constrained quadratic update problem, which preserves registered old responsibilities while reducing residuals with an effective safe response. We prove that reconstructive semantic protection weakly enlarges the safe-response operator relative to preserving an exact historical realization. Dynamic sufficiency and future-closure viability extend representation adequacy from current prediction to lawful future updating and continued learning. Growth Learning expands the lawful closure domain or lowers optimal closure cost without regression of the registered capability-cost frontier; a conditional commit rule maintains this order. Restricted-sector recoveries and a conditional representation theorem connect the framework to optimization, machine learning, and control. Together, these results organize adaptation, representation revision, and reusable capability within a common account of continued learning.
Sep 29, 2026

Continual Learning of Dynamical Systems in Recurrent Neural Networks through Recyclable Unit Gating

Dynamical Systems Reconstruction (DSR) aims to infer models from observed time series that reproduce a system's qualitative long-term behavior. Continual DSR (cDSR) requires learning new systems while preserving previously learned dynamics, yet even small parameter updates in recurrent models can qualitatively alter their behavior over long autonomous rollouts. We benchmark established continual learning (CL) methods spanning parameter regularization, replay, and parameter isolation on the fully trainable and interpretable Almost-Linear RNN (AL-RNN). Parameter isolation preserves earlier dynamics most effectively, but excessive task-specific allocations can rapidly exhaust a fixed-size network. We therefore introduce Continually-Recyclable Unit-Gating (CRUG), which conserves capacity through compact allocation and forward transfer. Differentiable gates trained with an L0L_0-based penalty select task-specific units, while unused units are recycled for subsequent tasks. Directed connections allow later tasks to reuse earlier representations without affecting the dynamics of previously committed units. CRUG achieves the strongest reconstruction--capacity trade-off among the tested methods with zero forgetting and reliably learns a heterogeneous sequence of nonlinear and chaotic systems. Furthermore, we show that forward transfer is more pronounced and useful when tasks share similar underlying dynamics. Lastly, we demonstrate that CRUG's advantages extend beyond autonomous cDSR to sequential cognitive tasks.
Sep 29, 2026

ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning

Multimodal continual instruction tuning (MCIT) aims to enable multimodal large language models to acquire new capabilities from sequential tasks while preserving previously learned knowledge. Existing methods primarily mitigate catastrophic forgetting by constraining parameter updates or separating task-specific adaptations. However, continual adaptation can also benefit from external knowledge that provides domain-specific information and reusable reasoning patterns for solving diverse instructions. For example, to answer "How many red cubes are to the left of the sphere?", domain knowledge can provide relevant concepts about objects and spatial relations, while reasoning knowledge can specify ordered operations such as object recognition, spatial filtering, and counting. Despite this potential, how to leverage external knowledge for continual adaptation remains largely unexplored in existing MCIT methods. To this end, we propose ReCAP, a retrieval-guided framework that leverages external knowledge to guide capability reuse during continual adaptation. At each continual stage, ReCAP uses external search and an LLM to incrementally build a knowledge base of domain, reasoning, and format knowledge based on the current-stage training data. For each instruction, retrieved domain knowledge guides generation, while retrieved reasoning knowledge selects and orders capability modules to form an instance-specific capability path. As these capability modules are reused across stages, subsequent adaptation can overwrite previously learned parameters. To enable stable cross-stage reuse, ReCAP introduces adaptive subspace recycling, which parameterizes reusable capability modules with shared bases and stage-specific cores, protects historically important directions while recycling residual capacity. Extensive experiments on MCIT benchmarks show that ReCAP achieves SOTA performance.
Sep 29, 2026

RAE-PPG: Duration-Grounded Retain-and-Extend Pretraining for PPG Foundation Models

Signal features derived from photoplethysmography (PPG) require different signal durations to characterize. Existing PPG foundation models treat duration as a pretraining or evaluation condition rather than using the different durations required by PPG features to organize self-supervision. We hypothesize that self-supervision should expand with signal duration, allowing a single encoder to progressively acquire additional features while preserving and reusing earlier learning. We introduce Retain-and-Extend PPG (RAE-PPG), which trains a single Transformer encoder successively on 10 s, 30 s, and 240 s inputs, adding supervision for signal features supported by each longer observation. The encoder is partitioned into duration-specific parameter groups, allowing later stages to reuse earlier groups while updating only the group assigned to the current stage. Selected earlier targets are reused to supervise later stages, encouraging the corresponding features to remain accessible in longer-input representations. Direct decoding from the final encoder shows that earlier features remain recoverable from longer-input representations, while later-stage features show higher mean decoding performance at their introduction durations. Controlled comparisons further show that prior-stage learning provides a better basis for learning newly introduced features at both transitions. Across 18 tasks from eight datasets, the final frozen encoder achieves the best observed score on 12 tasks compared with five existing PPG foundation models.
Sep 29, 2026

Generalizable Lifelong Model Editing via Preference Optimization

Knowledge editing enables rapid updates of specific factual knowledge in large language models (LLMs) without full retraining. However, more realistic scenarios call for a lifelong framework that handles continual updates rather than one-off modifications. In such settings, existing editing methods often overfit to target prompts, significantly degrading both the generalization of the edited knowledge and the model's general capabilities. To address this issue, we propose GLIME (Generalizable Lifelong Model Editing), which combines knowledge editing with preference optimization over generation behavior. GLIME further incorporates replay-based editing and a gradient constraint to preserve previously edited knowledge. Experimental results show that GLIME significantly improves knowledge generalization in lifelong editing settings while maintaining both editing performance and general capabilities.
Sep 28, 2026

Neural Succession: A Mesoscopic Theory of Invasion, Coexistence, and Stabilization in Continual Learning

Continual learning is usually studied through mechanisms that preserve old knowledge. We develop Successional Learning Theory (SLT), a mesoscopic account in which the current representation is a resident community, the incoming task is an invader, forgetting is resident displacement, joint retention is coexistence, replay is resident reinforcement, and training moves from establishment toward stabilization. Its empirical coordinate is directional pre-invasion compatibility, measured on the resident model before the incoming task is learned. Across eight experiments, compatibility orders later forgetting on the 20 directed Split-CIFAR-10 transitions (three-repeat r=-0.789, incoming-task cluster 95% CI [-0.90,-0.72], every repeat alone r<=-0.67), forecasts held-out forgetting with 24% lower error than a no-information baseline, and reproduces under controlled MNIST permutations and CIFAR-10 rotations (r=-0.804, -0.718). On an 84-transition suite, compatibility separates coexistence from exclusion at every retention threshold (AUC 0.93-0.97). Replay repairs every transition with at most 325 stored examples and is most efficient where displacement is largest. Compatibility reaches |r|=0.720, while activation, representation, Jacobian, and fixed-coefficient Lotka-Volterra specializations do not. Plasticity and feature turnover fall reliably from early to late training (15/15 and 14/15 runs). We formalize a minimum habitat-modification bound, a displacement floor, a sufficient coexistence condition, an identifiability law with a range-restriction corollary, successional stabilization, and local reinforcement. The identifiability law also predicts where the coordinate loses leverage, and the prediction matches three CIFAR-100 partitions and five optimizer regimes. SLT is a pre-adaptation diagnostic that complements replay, regularization, and projection methods.
Sep 28, 2026

Early Learning Shapes Later Directions Of Representation Change In Continual Learning

Representations continually change as a network learns new tasks. We ask whether early representational changes naturally form a geometric structure that continues to shape later learning. We identify a low-dimensional subspace of early representation drift, which we call a scaffold, and test whether it is reused across subsequent tasks. Across four pretrained visual encoders and two datasets, later representational changes consistently favor this early-defined subspace over matched random alternatives. This reuse is history-dependent: when networks experience different early tasks but identical later training inputs, each network preferentially reuses the scaffold induced by its own learning history. The same preference appears in individual optimizer updates, even though the network's dominant local response directions shift away from the original scaffold. Finally, constraining motion within the scaffold slows new-task acquisition more than matched random constraints, while effects on old-task retention are less consistent. In summary, these results suggest that early experience leaves a persistent geometric imprint on how neural networks adapt to future tasks.Code is available at https://github.com/YuantaoDeng/latent-scaffold.
Sep 28, 2026

SPACE-LoRA: Allocating Activation-Subspace Protection for Continual Learning

This study addresses the catastrophic forgetting problem that occurs when sequentially learning successive tasks using Low-Rank Adaptation (LoRA) from a lifelong learning perspective. While existing approaches have primarily constrained parameter updates or learning subspaces to reduce interference with past knowledge, they have not fully considered additive interference. This occurs when a newly added residual adapter on top of a fixed past model generates non-zero responses along input directions important for old tasks, thereby altering previous predictions. To this end, we propose Subspace Protection with Allocated Capacity for Efficient Continual Adaptation (SPACE-LoRA). SPACE-LoRA directly suppresses the responses of the new residual branch along input activation directions that are important for old tasks and adaptively determines the protection coverage for each module based on past-task sensitivity estimated via a common Fisher sensitivity-based coverage target. Under a fixed LoRA rank, this approach adaptively adjusts module-specific protection coverage while suppressing interference along input directions sensitive to old tasks. We assess the effectiveness of activation-subspace protection in mitigating catastrophic forgetting and examine the role of sensitivity-guided protection in continual learning across diverse tasks. Code is available at https://anonymous.4open.science/r/SPACE-LoRA-7864.
Sep 27, 2026

CARVE: Breaking Data Barriers in Chip Placement by Harnessing Reusable Expertise

Pretrained macro-placement policies can reduce repeated optimization across circuits, but deployment often exposes them to unfamiliar designs when the original training data are unavailable. Repeatedly fine-tuning a single serving model can overwrite earlier improvements, while simply saving checkpoints does not determine where they can be reliably reused. We introduce Continual Adaptation through the Reuse of Validated Expertise (CARVE), a framework that represents accumulated expertise as a frozen base policy, immutable specialists, and task-specific credentials obtained through local validation. For a new task, CARVE first checks existing specialists and trains a new specialist from the frozen base only when none qualifies. Under fixed task distributions and validation rules that control cumulative error, we establish expected-performance guarantees for repeated reuse. For bounded losses, we also derive matching worst-case bounds on the local samples needed for reliable reuse. In macro placement, a reuse-first follow-up reduces recorded training time by 58.5% (9.66 to 4.01 hours), while mean HPWL gain changes only from 8.41% to 7.86%. In a simulated receiving deployment, imported specialists are reused on six of seven new IBM circuits with no receiver-side training, achieving a 5.76% mean HPWL gain. Navigation studies provide complementary evidence on repair retention and repeated adaptation.
Sep 26, 2026

When the Merge Coefficient Stops Mattering: Proximity Regularized Merging for Continual LoRA Adaptation

Rehearsal-free continual learning with parameter-efficient adapters can be cast as a sequence of task-vector write-in operations: for each new task, a low-rank adapter is learned and merged into a running model. We propose Proximity Regularized Merging (PRM), a minimal modification to sequential LoRA merging that adds a proximal penalty during task-vector training without changing the subsequent write-in rule. PRM acts as a robust task-vector regularizer: in the reported Base->+Prox diagnostics, it improves AAA across multiple write-in rules, backbones, and class-incremental settings, while its fixed-coefficient variant remains competitive with strong coefficient-based baselines. Mechanistically, matched-prefix norm controls and proximal-strength sweeps show that proximal training shrinks the task-vector radius, lowers Fisher-weighted interference, broadens the coefficient plateau, and exposes a stability-plasticity trade-off. Together, these results suggest that the effectiveness of sequential LoRA merging depends not only on how much of a task vector is written in, but also on whether the task vector itself has been trained to be mergeable.
Sep 26, 2026

What Should We Freeze? Guarded Freezing: Connectivity Shapes the Fine-Tuning of Pretrained Models

When adapting pre-trained models through fine-tuning, freezing weights alone might not preserve performance, as updates elsewhere can change the inputs to the frozen core, ultimately affecting overall performance. We first analyse the case where a selected frozen core can be isolated and propose removal-value, a capacity-based score that approximates HOPE's removal cost averaged over removal orders. We show that in VGG-8, cutting paths from trainable neurons into a frozen core makes selection using this score useful: 70%70\% frozen preserves 5.22±0.515.22\pm0.51 percentage points more old-task accuracy than DEFT at similar new-task accuracy. In transformers, shared residual streams leave paths into frozen neurons open. For this case, we derive drift-value, a forward-only proxy for the output disturbance from updating each weight entry under a local update model. In language models, at 40 epochs, this policy exceeds adapted Wanda and RIA freezing scores in settings with substantial retention loss, while its differences from Fisher remain unresolved. After 160 epochs on Qwen2.5-1.5B, it retains 0.0433±0.01020.0433\pm0.0102 more than static Fisher. In DINOv3 vision-transformer adaptation to point clouds, drift-value retains 0.4400.440 image accuracy versus 0.1870.187 for a random mask of the same count. These results motivate Guarded Freezing: select by removal-value when incoming paths are cut, and by drift-value when they remain.
Sep 24, 2026

Intrinsic-Extrinsic Coupling in Learning Dynamics

A learner's current observations need not determine its response to further training. We formulate intrinsic-extrinsic coupling through the continuation-conditioned value of a constrained learning-state intervention, with observation-relative fibers describing present agreement. An executable finite-frame classifier-head write protects current logits while repairing specified historical margins under finite-precision acceptance checks. We distinguish local admissibility, continuation-conditioned intervention value, and complete-policy performance. A matched four-cell contrast identifies readout-specific non-additivity between the same intrinsic intervention and alternative external continuations. In a CLINC-derived class-incremental setting, replay changes the write's 32-update contribution from five correct predictions to zero. Nonzero interactions also occur under output distillation, with a RoBERTa backbone, and under optimizer-native SGDW dynamics. Under SGDW, correct-count interactions are negative in all three activated roots at 128 updates, showing that coupling need not imply positive synergy. The mathematical analysis distinguishes feasible local repairs and favorable terminal outputs from training-reachable repair regions. Separate coordination tests show that content controls match or exceed the development gain, while a five-root fresh-test comparison with Fiber present in every arm shows root-dependent rather than uniformly beneficial correct-count effects. On the secondary cross-entropy readout, guided allocation yields lower mean loss than standard replay in all five pairs. Together, these results make intrinsic-extrinsic coupling operational by connecting executable state geometry to continuation-conditioned value, matched interaction identification, and closed-loop coordination, while separating identified coupling from complete-policy performance.