# transfer learning

#graph neural networks#node classification#link prediction#transfer learning

Same Graph Cross-Task Transfer in GNNs: Protocols and Predictors

Neelam Akula, Surbhi Kumar, Murat Kantarcioglu +1

The paper defines a clean evaluation protocol for transferring knowledge between node classification and link prediction on the same graph, shows that transfer is directionally dep…

cs.AI2026

PerturbMap: Cross-Context Transfer of Single-Cell Perturbation Responses

Panpan Cui, Yiqi Liu, Wenhao Sun

PerturbMap predicts missing single‑cell perturbation effects in a new cellular context by combining a low‑rank local model with ridge‑based transfer of measured responses from sour…

#single-cell perturbation#transfer learning#cross-context prediction#low-rank modeling

quant-ph2026

Quantum machine learning interatomic potential: Application of variational quantum algorithm

Kohei Numata, Wataru Mizukami, Kosuke Mitarai +2

The paper integrates a variational quantum circuit into a classical neural network for interatomic potentials, retraining the ANI model via quantum transfer learning and showing mo…

#quantum machine learning#interatomic potentials#variational quantum algorithms#transfer learning

#pathology foundation models#mitotic figure detection#dense object detection#self-supervised learning

Beyond Classification: Pathology Foundation Models as Detection Encoders for Mitotic Figures

Sweta Banerjee, Alireza Teimoury, Nils Porsche +11

The paper evaluates whether pathology foundation models can serve as effective backbones for dense detection of mitotic figures, comparing several self‑supervised models to a ResNe…

#acne severity grading#transfer learning#efficientnet#grad-cam

Interpretable Image-Level Acne Severity Grading via EfficientNet-B0 Transfer Learning and Grad-CAM

Sophie Zeng, Sean Kalaycioglu, Collin Hong +1

The paper introduces a four‑class acne severity grading model that uses transfer learning with an EfficientNet‑B0 backbone and Grad‑CAM visualizations to achieve high accuracy and…

#graph neural networks#foundation models#agent-based learning#information flow control

AgentGFM: A Graph Foundation Model with Node-Agent Information-Flow Control

Jingbo Cui, Jitao Zhao, Di Jin +1

The paper introduces AgentGFM, a graph foundation model where each node acts as an agent that autonomously decides how to propagate information using a trainable policy, enabling a…

#lung disease classification#chest x-ray#deep learning#transfer learning

Classification of Disease from Lungs X-ray Images using VGG16, VGG19 and ResNet50 Models

Nand Lal Yadav, Rajesh Kumar, Satyendra Singh +1

The paper evaluates deep convolutional neural networks (VGG16, VGG19, and ResNet50) for classifying lung diseases such as pneumonia, tuberculosis, and lung cancer from chest X‑ray…

cs.AI2026

Rethinking Self-Evolution: A Constrained Exploration-Exploitation Process for Mitigating Skill Overfitting

Hongqiang Lin, Chao Liu, Xiaofan Bai +4

The paper introduces SkillBoost, a three-stage framework that reduces overfitting of trainable skills in large language model agents by balancing constrained exploitation of failur…

#large language models#skill learning#exploration-exploitation#overfitting mitigation

#ct imaging#vision-language models#anatomy segmentation#transfer learning

Anatomy Contextualized Adaption of CT Foundation Models

Roshan Kenia, Stephanie L McNamara, William Lotter

The paper proposes Anatomy Contextualized Adaptation (ACA), a lightweight method that adapts frozen CT vision-language foundation models to align anatomy-level visual features with…

#concept drift#hidden markov model#weight trajectory analysis#transfer learning

Latent States in Neural Networks: Recovering the Temporal Structure of Drifting Data from Model Weights

Kevin Guan

The paper studies whether hidden regimes in temporally drifting data streams can be identified from the weights of sequentially trained neural classifiers by fitting a hidden Marko…

cs.SD2026

Does EEG Foundation Models Transfer to Speech? A Benchmark on Overt and Imagined Speech Decoding

Owais Mujtaba Khanday, Mohamed Baha Ben Ticha, Sanae Belfrouh +2

The paper benchmarks two EEG foundation models on overt and imagined speech decoding tasks, comparing them to established convolutional baselines, and finds that large‑scale EEG pr…

#eeg foundation models#speech decoding#overt speech#imagined speech

#transfer learning#cross-configuration learning#legacy data reuse#robot manipulation

When Does Legacy Data Start to Help? Emergent Transfer in Cross-Configuration Robot Learning

Tao Wang, Hudson Hou, Yingdong Hu +7

The paper investigates when demonstration data collected on an older robot configuration becomes useful for training a newer robot, revealing a three‑phase pattern where legacy dat…

#low-rank adaptation#lora initialization#gradient preconditioning#transfer learning

Between Gradient and Natural Gradient: A Continuum of LoRA Initializations

Dianze Liu, Farshid Ghezelbash

The paper introduces Unified LoRA (ULoRA), a two-parameter family of preconditioned gradient initializations for low‑rank adaptation, showing that optimal initialization varies by…

eess.SY2026

GenTL: A General Transfer Learning Model for Building Thermal Dynamics

Fabian Raisch, Thomas Krug, Christoph Goebel +1

The paper introduces GenTL, a pretrained LSTM model that can be fine‑tuned for many single‑family houses, removing the need for source‑building selection and cutting prediction err…

#transfer learning#building thermal dynamics#LSTM#energy prediction

#plasticity#continual learning#transfer learning#information theory

Local Redundancy: An Information-Theoretic Measure of Plasticity from Synthetic Memorization

Jiaxuan Cheng

The paper introduces local redundancy, an information‑theoretic measure of neural network plasticity derived from universal compression theory, and shows that a computable lower bo…

#vision transformers#layer fusion#linear probing#transfer learning

Attentive multilayer fusion for vision transformers

Laure Ciernik, Marco Morik, Lukas Thede +4

The paper introduces Attentive Layer Fusion (ALF), a method that dynamically combines representations from all layers of a Vision Transformer to improve linear probing on downstrea…

#test-time learning#meta-learning#language agents#adaptation policies

Learning to Learn-at-Test-Time: Language Agents with Learnable Adaptation Policies

Zhanzhi Lou, Hui Chen, Yibo Li +2

The paper introduces Meta-TTL, a bi‑level optimization framework that learns adaptation policies for test‑time learning in language agents, using evolutionary search to improve per…

physics.geo-ph2026

2.5D Transformer: An Efficient 3D Seismic Interpolation Method without Full 3D Training

Changxin Wei, Xintong Dong, Xinyang Wang

The paper introduces a 2.5‑dimensional Transformer that leverages 2D Transformer encoders and specialized adapters to interpolate 3D seismic data efficiently, using a two‑stage tra…

#seismic interpolation#transformer models#transfer learning#3d data processing

#vehicle dynamics#model predictive control#kinodynamic prediction#terrain adaptation

Adapting Generalist Vehicle Models for High-Speed MPC Across Terrains

Rwik Rana, Jesse Quattrociocchi, Christian Ellis +3

The paper introduces OptCar, a method for adapting a generalist forward kinodynamic model to a specific vehicle using minimal real-world data and synthetic rollouts, improving high…

#large language models#post-training#proxy models#reinforcement learning

Proxy OPD: On-Policy Distillation with Transferable Relative Proxy Update

Daocheng Fu, Rong Wu, Yu Yang +7

The paper introduces PUST, a framework that uses a lightweight proxy model to explore high‑reward behaviors and then transfers the relative improvement signals to a larger primary…

#task inversion#imitation learning#skill generalization#transfer learning

Task Parameter Extrapolation via Learning Inverse Tasks from Forward Demonstrations

Serdar Bahar, Fatih Dogangun, Matteo Saveriano +2

The paper introduces a joint learning framework that builds a shared representation of forward and inverse tasks, enabling robots to infer and execute inverse tasks from only forwa…

#safe navigation#goal‑conditioned planning#stochastic path prior#transfer learning

AutoPath: Learning Transferable Goal-Conditioned Stochastic Path Prior for Safe Navigation Without Human Demonstrations

Ziyang Zhang, Boyang Zhou, Zesong Yang +8

The paper proposes a goal‑conditioned stochastic path prior that learns a transferable distribution over local navigation paths from limited observations, enabling safe, multimodal…