AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

Tail Cues, Principal Corrections: Plug-and-Play Rectification for Open-Set Test-Time Adaptation

Yingkai Yang, Chaoqi Chen, Hui Huang

NeurIPS 2026 · Label-efficient and robust learning › Test-time adaptation · approach: Realistic and stable test-time adaptation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes TALOR, a lightweight backward-free plug-and-play rectification module for open-set test-time adaptation that uses the source linear head as a geometric anchor.

Problem

Open-set test-time adaptation on non-stationary streams where covariate and semantic shifts coexist; existing TTA methods update parameters or maintain mutable memories, which is costly and vulnerable to unknown samples.

Approach

A backward-free rectification module that uses the source linear head as a geometric anchor to correct predictions, avoiding parameter updates and mutable memory banks.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add TALOR as a plug-in on the frozen source modelSource ViT-B/16 · ImageNet-C open-set (mean H-score)H-score: 56.4 -> 66.1Improved
add TALOR post-correction to COMECOME · ImageNet-C open-set (mean H-score)H-score: 63.7 -> 68.4Improved
add TALOR to SARSAR · ImageNet-C open-set (mean H-score)H-score: 51.1 -> 65.4Improved
add TALOR to EATAEATA · Textures-C AUCAUC: -1.7 AUC on Textures-C; +5.2 mean H-scoreMixed
add TALOR to base methods on LAION-CCoTTA, EATA, OSTTA, ViDA · LAION-C mean H-scoreH-score: +10.8, +7.4, +11.5, +10.7Improved
add TALOR in closed-set protocolsstrong baselines such as DPCore · closed-set LAION-C accuracyaccuracy: 65.3 (+0.1)Improved
use only the principal mean term (remove tail slope) (ablation)TALOR on Source · ImageNet-C open-set meanH-score: 66.1 -> 62.3Hurt
use only the tail-to-principal slope term (remove mean) (ablation)TALOR on Source · ImageNet-C open-set meanH-score: 66.1 -> 61.4Hurt
add soft ID/OOD routing instead of one affine rectifier (ablation)TALOR on Source · ImageNet-C open-set meanH-score: 65.4 -> 66.1Improved
use ground-truth routing instead of unsupervised GMM routing (ablation)TALOR on Source · ImageNet-C open-set meanH-score: 66.1 -> 66.9Improved
vary SVD energy threshold tau (ablation)TALOR · ImageNet-C open-setH-scoreNo clear effect
remove EMA smoothing (very small beta) (ablation)TALOR · ImageNet-C open-setH-scoreHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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