AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

LANE: Label-Aware Noise Elimination for Fine-Grained Text Classification

Tiberiu Sosea, Cornelia Caragea

ICLR 2026 · Label-efficient and robust learning › Noisy labels · approach: Instance-dependent noise modeling

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 label-aware margin (ALM) with semantic weights instead of plain AUM weighting (ablation)LANE · RCV1 (20% noise)micro F1: a boost of 4.2% over LANE−simImproved
add weighted cross entropy based on ALM (ablation)LANE · RCV1 (20% noise)micro F1: 3.2% over LANE−almImproved
replace hard removal of low-AUM samples with LANE soft weighting (ablation)AUM · RCV1 (20% noise)micro F1: 1.8% over AUMImproved
use LANE instead of AUM (ablation)AUM · SciHTC (original labels)F1: outperforming AUM by 2.9% and LANE−sim by 1.7%Improved
infer Gaussian threshold dynamically instead of fixed mu=0 (ablation)LANE · Yelpaccuracy: 64.4 -> 65.2Improved
infer Gaussian threshold dynamically instead of fixed mu=-1 (ablation)LANE · Yelpaccuracy: 64.1 -> 65.2Improved
infer Gaussian threshold dynamically instead of fixed mu=0 (ablation)LANE · Yahooaccuracy: 68.1 -> 68.9Improved
infer Gaussian threshold dynamically instead of fixed mu=-1 (ablation)LANE · Yahooaccuracy: 65.1 -> 68.9Improved
swap BERT backbone for RoBERTaLANE · GoEmotionswF1: 66.5 -> 66.8Improved
add LANE on top of RoBERTa baselineRoBERTa · Yahoo Answeraccuracy: 75.7 -> 78.9Improved
use LANE rather than DivideMixDivideMix · Yahoo Answeraccuracy: 69.17 -> 78.4Improved
use LANE rather than few-shot Llama-2Llama-2 13B · SciHTC (20% noise)macro F1: 15.1 -> 30.5Improved

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