AI Paper Atlas

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

NBCG: Nash-Bargained Causal Game for Long-Tailed Multi-Label NLP

Jing Yang, Jusheng Zhang, Keze Wang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive learning for imbalanced data

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
replace Nash bargaining objective with a static weighted sum of coalition losses (ablation)NBCG · DBpedia Rare-F1 @30%Rare-label F1: 61.8 -> 65.8Improved
add Neural SEM causal coalitions instead of random label partitions (ablation)NBCG · DBpedia Rare-F1 @30%Rare-label F1: 63.5 -> 65.8Improved
add adaptive (EMA) disagreement point instead of fixed dk = 0 (ablation)NBCG · DBpedia Rare-F1 @30%Rare-label F1: 64.2 -> 65.8Improved
increase number of players (coalitions) from 1 to 5 (ablation)NBCG · 20 NewsgroupsmAP: 88.50% -> 92.87%Improved
increase number of players beyond 5 (to 10) (ablation)NBCG · 20 NewsgroupsmAP: 92.87% -> 90.80%Hurt
use NBCG instead of CCG on high label cardinality samplesCCG · Reuters mAP, cardinality >=6mAP: 73.1 -> 78.5Improved
use NBCG instead of CCG on low cardinality samplesCCG · Reuters mAP, cardinality 1-2mAP: 89.2 -> 89.8Improved
use NBCG instead of RoBERTa fine-tuning for rare labelsRoBERTa · DBpedia Rare-F1 @30%Rare-label F1: 55.4 -> 65.8Improved
add NBCG training phases (Neural SEM plus Nash optimization)RoBERTa (Base) · Reuters training timerelative train time: 1.00x -> 1.15xHurt
add NBCG to baseline training, head-label performanceRoBERTa · Reuters head labelslabel-wise F1No clear effect

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