AI Paper Atlas

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

CORE-MTL: Rethinking Gradient Balancing via Causal Orthogonal Representations

Chengfeng Wu, Tao Zou, Yanru Wu, Jingge Wang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 0 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 dual-stream semantic-residual backbone (ablation)Vanilla MTL · NYUv2mIoU: 0.5249 -> 0.5352Improved
add reconstruction grounding to dual stream (ablation)DS · NYUv2mIoU: 0.5352 -> 0.5424Improved
add CKA independence loss (ablation)DS+Grounding · NYUv2mIoU: 0.5424 -> 0.5547Improved
add counterfactual augmentation (ablation)DS+Grounding+CKA · NYUv2mIoU: 0.5547 -> 0.5693Improved
add CKA independence loss (ablation)DS+Grounding · NYUv2 surface normal11.25 deg accuracy: 0.3517 -> 0.3165Hurt
use physics-based hard grounding decoder instead of generic soft decoder (ablation)CORE-MTL · NYUv2mIoU: 0.5652 -> 0.5693Improved
use physics-based hard grounding decoder instead of generic soft decoder (ablation)CORE-MTL · NYUv2 depthRel Err: 0.1427 -> 0.1466Hurt
set lambda_CKA to 1.0 instead of 0 (ablation)CORE-MTL · NYUv2mIoU: 0.5636 -> 0.5694Improved
raise lambda_CKA to 100.0 from 1.0 (ablation)CORE-MTL · NYUv2mIoU: 0.5694 -> 0.5551Hurt
use CORE-MTL instead of RepMTLCORE-MTL · Cityscapes depth Rel ErrRel Err: 44.3230 -> 19.6088Improved
use CORE-MTL instead of Equal WeightingCORE-MTL · Cityscapes-CRel Err: 58.75 -> 38.10Improved
use CORE-MTL instead of Equal Weighting on 10-task CelebACORE-MTL · CelebA 10 tasksAvg accuracy (%): 88.52 -> 89.39Improved

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