Domain Adaptation and Few-Shot Learning
New papers in domain adaptation and few-shot learning, the code and data they release, and the open problems their authors admit. Rebuilt every day from OpenAlex and arXiv.
New papers
MASS: Multi-mask Selective Self-Distillation for Leakage-Resilient Reasoning
We propose Multi-Mask Selective Self-Distillation (MASS), which approximates the unavailable marginal over diverse reference realizations using multiple independently masked views of one verified solution.
In-Context Learning for Multi-Fidelity Multiscale Model Closure and Cross-Operator Out-Of-Distribution Generalization
In this work, we propose an in-context multi-fidelity learning framework for learning multiple multi-fidelity operators with one shared model.
Variational Consequence-Driven Offline Reinforcement Learning
To bypass this issue, we propose Structural Behavior Regularization (StBR), which replaces the intractable objective with a closed-form latent divergence.
Learning a Prior-Data Fitted Network for Graph Foundation Models via Structural Causal Model Priors
To address these challenges, we propose **GSCM-PFN** (Graph SCM-based Prior-Data Fitted Network), a graph PFN pre-trained on synthetic data generated from an SCM-based prior that jointly models nodes and graph topology.
Continual Adaptation Across Modalities: Learning and Forgetting in Unified Multimodal Models
In this paper, we systematically investigate how architectural design and adaptation method shape intra-modal learning and cross-modal interference between the generation and understanding sides.
Dual DeltaNet: Value-Side Memory Reorganization for Linear Attention
Based on this observation, we propose Dual DeltaNet (DDN), which extends Kimi Delta Attention (KDA) with a two-sided delta rule.
Loss is Not Behavior: A Unified Output-Space Analysis of Gradient-Based Machine Unlearning
Our framework yields precomputable leverage diagnostics for auditing forget sets to identify what type of gradient-based unlearning is likely to be successful.
Towards Reliable, Scalable, and Value-Aligned Generative Recommendation for Large-Scale Advertising
We present GR4AD-X, which addresses these limitations through reliable SID construction, serving-efficient capacity scaling, and unified value alignment.
Knowledge-structured and evidence-aware semantic self-distillation for glomerular pathology foundation model
Here, we introduce GloSAGE, a semantic self-distillation framework that uses lesion profiles predicted by a pretrained visual teacher to construct structured pathology descriptions.
Detecting Fine-Grained Differences between Image Datasets
Additionally, by designing a new benchmark that focuses on low-prevalence and multi-concept distinctions, we find that selecting nonredundant concepts instead of only the most distinguishing concepts matters, and we introduce techniques for how to effectively do this.
Cross-Primitive Consistency in Jev
This work introduces PrimaCons, a four-axis benchmark for assessing cross-primitive consistency while preserving the semantics of the query.
BridgeRouter: Gain-Aware Instance Routing over Cross-Modal Bridge Portfolios
Lightweight representation bridges make pretrained, task-adapted EEG foundation models executable on other biosignal modalities without retraining their backbones.
Prototype-Guided Dynamic Domain Modulation and a Large-Scale Multi-SKU Benchmark for Shampoo Bottle Surface Defect Detection
We introduce Bottle-SKU, a large-scale multi-SKU benchmark for shampoo bottle surface defect detection.
DES-JEPA: Learning a Shared Vocabulary for High-Dimensional Discrete Event Sequences
To address this, we propose two complementary modules: (i) DES-JEPA, a joint-embedding predictive architecture that learns a shared discrete vocabulary from synthetic stochastic processes with unaligned event vocabularies; and (ii) DES-PFN, a discrete event sequence prior-fitted network that…
SaiVLA-JEPA: Evaluating Future Latent Prediction in JEPA-Conditioned VLA Control
Joint-embedding predictive architectures (JEPAs) learn to predict representations rather than pixels, offering a compact way to model how visual scenes evolve.
PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation
We propose PT$^3$, a prototype-guided self-supervised pretraining framework for mmWave human point clouds.
Episodic looking - using reinforcement learning to bias perception enhances few-shot and continual learning with protection from forgetting
This paper proposes a solution based on the Complementary Learning System architecture (CLS), in which a Short-Term Memory (STM) delegates generalization to a slow, statistical Long-Term Memory (LTM).
inkGround: Benchmarking Fine-grained Visual Grounding Capabilities of Vision-Language Models for Cursive Handwriting in the Wild
In this work, we introduce inkGround, a benchmark for stroke trajectory recovery from full-page, camera-captured handwriting. inkGround pairs 537 page images (~18.7k words, ~163k strokes) with reference trajectories produced by a pretrained offline-to-online model, filtered by OCR agreement, and…
Least decision information-based dual-anchor framework for cross-domain fault diagnosis
In this study, a new concept termed ε -least decision information ( ε -LDI) is proposed to characterize the essential decision-making information of cross-domain diagnostic models.
The Architecture of Permanence: Deterministic Edge Governance Across Heterogeneous NVIDIA Robotics on Commodity Silicon
Executive Overview & Problem Context Current Vision-Language-Action (VLA) foundation models rely heavily on brute-force scale, empirical loss minimization, and post-hoc alignment (such as RLHF).
Released code and data
Dual DeltaNet: Value-Side Memory Reorganization for Linear Attention
Compared with KDA, GDN-2, and Mamba-3, DDN achieves the strongest overall performance across language modeling, zero-shot downstream tasks, and retrieval, while incurring modest training and inference overhead relative to KDA. Code: https://github.com/ReichtumQian/DualDeltaNet
Detecting Fine-Grained Differences between Image Datasets
Additionally, by designing a new benchmark that focuses on low-prevalence and multi-concept distinctions, we find that selecting nonredundant concepts instead of only the most distinguishing concepts matters, and we introduce techniques for how to effectively do this.
Cross-Primitive Consistency in Jev
Code and benchmark is available at https://github.com/67L1/PrimaCons-CRPR.
Prototype-Guided Dynamic Domain Modulation and a Large-Scale Multi-SKU Benchmark for Shampoo Bottle Surface Defect Detection
We introduce Bottle-SKU, a large-scale multi-SKU benchmark for shampoo bottle surface defect detection.
Learning from Points and Patches: A Pre-trained Hybrid Attention Transformer for Time Series Analysis
Code is available at https://github.com/TangPeiwang/TimeHAT.
Nonlinear Direct Feedback Alignment for Scalable Backpropagation-Free Training
The code is available at https://github.com/YanFan321/NDFA.
LLM knowledge for improved Out-of-Distribution Tabular Prediction - A Cross-Validated Regularisation Approach
Code, priors and results: https://github.com/hugohammer/llm-priors-ood.
SimplexLoRA: Dynamic Rank Updating via Alternating Minimization with Simplex Projection
Our code is available at https://github.com/brain-lab-research/SimplexLoRA.
Can Complementary Signals Bridge Similarity Islands? Manifold-Augmented Graph Embedding for Multimodal Recommendation
Our code is available at https://github.com/Alex-bupt/mage.
PRISM: Pathology Reliability In Scarce-label Medicine. A Tile-level Benchmark of Pathology Foundation Models
PRISM is released as a pip-installable package with reference results for all 288 cells. Code and data: https://github.com/alicemkyn/prism-benchmark
Open problems authors admit
Episodic looking - using reinforcement learning to bias perception enhances few-shot and continual learning with protection from forgetting
In future work, we will seek to address this by replaying STM memories to the LTM.
The Architecture of Permanence: Deterministic Edge Governance Across Heterogeneous NVIDIA Robotics on Commodity Silicon
Future Work: Plans focus on expanding topological invariance to high-degree-of-freedom humanoid systems (e.g., NVIDIA Project GROOT), extending the invariant anchor manifold across 20+ simultaneous joint limits, whole-body center-of-mass (CoM) stability polygons, and zero-moment point (ZMP) dynamic constraints.
A Priori Domain Shift Index for Assessing Transferability in Federated Fault Diagnosis of Rotating Machinery
The index is therefore best regarded as a privacy-friendly screening dissimilarity rather than a quantitative predictor of accuracy loss; validation on industrial field data remains future work.
CaLLM: Continual Adaptation of LLMs via Non-Parametric Adapter Routing
Continual adaptation of large language models (LLMs) to evolving task distributions while mitigating catastrophic forgetting remains an open problem.
Progressively Refining SAM for Accurate Phytoplankton Segmentation
In addition, we construct a phytoplankton segmentation dataset by manually annotating instance-level masks for an existing dataset, providing a valuable resource for future research.
R obot L earning M echanisms and F rontier P aradigms u nder the D ual C hallenges of F ew- S hot L earning and S parse R ewards
In addition, this essay also constructs a systematic taxonomy comparison framework and looks forward to future research directions such as security exploration and embodied alignment.
Shared-first risk-aware open-set domain adaptation for catalogue-absent structural state identification
The evidence is simulation based; real-bridge and long-term field validation are left for future work.
Rescuing Your Pruning: Rethinking Visual Token Pruning and Post-Training Recovery in MLLMs
We hope these findings offer insights for future research on joint visual token compression.
Adaptive temperature distillation with curriculum learning for lightweight multilingual language teaching models
It should be noted that the current framework is evaluated exclusively on discriminative tasks; its applicability to generative tasks such as machine translation and abstractive summarization remains an avenue for future work.
Multimodal continual federated learning for human activity recognition: Trends and challenges
Finally, this review outlines research opportunities for multimodal federated continual HAR, identifies unresolved trade-offs among privacy, adaptability, and scalability, and suggests future directions, such as standardized benchmarks and interdisciplinary collaboration, to bridge laboratory settings and real-world deployment.