AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Rep-MTL: Unleashing the Power of Representation-level Task Saliency for Multi-Task Learning

Zedong Wang, Siyuan Li, Dan Xu

ICCV 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add cross-task saliency alignment (CSA) to EW (ablation)Rep-MTL · NYUv2task-level gain: -1.78 -> +1.06Improved
add task-specific saliency regulation (TSR) to EW (ablation)Rep-MTL · NYUv2task-level gain: -1.78 -> +0.23Improved
add TSR on top of CSA (ablation)Rep-MTL · NYUv2task-level gain: +1.06 -> +1.70Improved
add CSA on top of TSR (ablation)Rep-MTL · NYUv2task-level gain: +0.23 -> +1.70Improved
add CSA to EW (ablation)Rep-MTL · Cityscapestask-level gain: -2.05 -> +0.21Improved
add TSR to EW alone (ablation)Rep-MTL · Cityscapestask-level gain: -2.05 -> -0.34Improved
add both CSA and TSR to EW (ablation)Rep-MTL · Cityscapestask-level gain: -2.05 -> +0.62Improved
use Rep-MTL regularization instead of plain EWEW · NYUv2task-level gain: -1.78 -> +1.70Improved
use Rep-MTL instead of DB-MTLDB-MTL · NYUv2task-level gain: +1.15 -> +1.70Improved
use Rep-MTL instead of DB-MTLDB-MTL · Cityscapestask-level gain: +0.20 -> +0.62Improved
use Rep-MTL instead of EWEW · Office-Homeaverage accuracy: 78.17 -> 79.14Improved
use Rep-MTL instead of EWEW · Office-31task-level gain: -0.61 -> +1.31Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy