Task Transferability Estimation

Follow this topic Ask about this topic

Task transferability estimation asks whether knowledge from a source model, dataset, task, or skill will improve performance on a target, helping choose what to reuse before expensive training or deployment. Methods compare frozen representations using embedding distances, CKA, or topological features; measure gradient or Fisher alignment; or learn rankers from task metadata and small transfer probes. Applications span vision and medical segmentation, cross-lingual NLP, and reinforcement learning, where estimates can route among models or skills, compose prior policies, or gate transfer against target-only baselines. Results show useful predictive signals, but rankings can change with target samples, metrics, architecture, and task type, while similarity may miss negative or directional transfer; reliable calibration and validation remain open challenges.

Latest papers

38 papers11 papers in the last four weeks, up 83% on the four weeks before. 0.1% of all new papers.

PAS: Estimating the target accuracy before domain adaptation
Apr 10, 2026Raphaella Diniz, Jackson de Faria, Martin EsterDomain Adaptation Task Transferability Estimation
School of Computing Science Simon Fraser University Canada
Information-Theoretic Requirements for Gradient-Based Task Affinity Estimation in Multi-Task Learning
Apr 9, 2026Jasper Zhang, Bryan ChengMulti-Task Learning Task Transferability Estimation
Great Neck South High School
Cross-Domain Policy Optimization via Bellman Consistency and Hybrid Critics
Mar 12, 2026Ming-Hong Chen, Kuan-Chen Pan, You-De Huang +2Reinforcement Learning Cross-Domain Transfer Learning
National Yang Ming Chiao Tung University, Hsinchu, Taiwan · Applied Machine Learning, Meta AI, Menlo Park, CA, USA
SAE as a Crystal Ball: Interpretable Features Predict Cross-domain Transferability of LLMs without Training
Mar 3, 2026Qi Zhang, Yifei Wang, Xiaohan Wang +4Task Transferability Estimation Cross-Domain Transfer Learning
State Key Lab of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University · Meituan · Amazon AGI SF Lab +1
Topology-Driven Transferability Estimation of Medical Foundation Models for Segmentation
Feb 27, 2026Jiaqi Tang, Shaoyang Zhang, Xiaoqi Wang +3Task Transferability Estimation Medical Imaging Foundation Models
Peking University, Beijing, China · National Institute of Health Data Science, Peking University, Beijing, China · Institute of Medical Technology, Peking University, Beijing, China +3
Ensemble Prediction of Task Affinity for Efficient Multi-Task Learning
Feb 20, 2026Afiya Ayman, Ayan Mukhopadhyay, Aron LaszkaMulti-Task Learning Task Transferability Estimation
Pennsylvania State University · College of William & Mary
Model Specific Task Similarity for Vision Language Model Selection via Layer Conductance
Feb 1, 2026Wei Yang, Hong Xie, Tao Tan +3VLM Evaluation Vision-Language Models
School of AI and Data Science USTC
Taskonomy: Disentangling Task Transfer Learning
Apr 23, 2018Amir Zamir, Alexander Sax, William Shen +3Transfer Learning Task Transferability Estimation
Stanford University · University of California, Berkeley

Apr 10, 2026J/K move · Enter open · S save

PAS: Estimating the target accuracy before domain adaptation

Raphaella Diniz, Jackson de Faria, Martin Ester

School of Computing Science Simon Fraser University Canada

Domain Adaptation Task Transferability Estimation

Open paper AskPDF arXiv

The goal of domain adaptation is to make predictions for unlabeled samples from a target domain with the help of labeled samples from a different but related source domain. The performance of domain adaptation methods is highly influenced by the choice of source domain and pre-trained feature extractor. However, the selection of source data and pre-trained model is not trivial due to the absence of a labeled validation set for the target domain and the large number of available pre-trained models. In this work, we propose PAS, a novel score designed to estimate the transferability of a source domain set and a pre-trained feature extractor to a target classification task before actually performing domain adaptation. PAS leverages the generalization power of pre-trained models and assesses source-target compatibility based on the pre-trained feature embeddings. We integrate PAS into a framework that indicates the most relevant pre-trained model and source domain among multiple candidates, thus improving target accuracy while reducing the computational overhead. Extensive experiments on image classification benchmarks demonstrate that PAS correlates strongly with actual target accuracy and consistently guides the selection of the best-performing pre-trained model and source domain for adaptation.