Challenge: Current methods for interpolative data augmentation select samples at random, which might make it difficult for the model to generalize better and converge faster.
Approach: They propose a curriculum-based learning method that leverages the relative position of samples in hyperbolic embedding space as a complexity measure to gradually mix up increasingly difficult and diverse samples along training.
Outcome: The proposed method achieves state-of-the-art results over existing methods on 10 benchmark datasets across 4 languages in text classification and named-entity recognition tasks.

Similar Papers

Mixup-Transformer: Dynamic Data Augmentation for NLP Tasks (2020.coling-main)

Copied to clipboard

Challenge: Recent work on data augmentation techniques that interpolate inputs and labels shows strong effectiveness in image classification.
Approach: They propose to integrate mixup to transformer-based pre-trained architecture for NLP tasks while keeping the whole end-to-end training system.
Outcome: The proposed framework improves on GLUEbenchmark and transformer-based learning models while keeping the whole end-to-end training system.
DMix: Adaptive Distance-aware Interpolative Mixup (2022.acl-short)

Copied to clipboard

Challenge: Interpolation-based regularisation methods such as Mixup have shown to be effective for various tasks and modalities.
Approach: They propose an adaptive distance-aware interpolative Mixup that selects samples based on their diversity in the embedding space.
Outcome: The proposed method achieves state-of-the-art on sentence classification over existing methods on 8 benchmark datasets across English, Arabic, Turkish, and Hindi languages while achieving benchmark F1 scores in 3 times less number of iterations.
Beyond Random Sampling: Efficient Language Model Pretraining via Curriculum Learning (2026.eacl-long)

Copied to clipboard

Challenge: Curriculum learning has improved efficiency across machine learning domains, but remains underexplored for language model pretraining.
Approach: They present a systematic investigation of curriculum learning in LLM pretraining . they use vanilla curriculum learning, pacing-based sampling, and interleaved curricula .
Outcome: The proposed framework accelerates convergence in early and mid-training phases, reducing training steps by 18-45% to reach baseline performance.
HuCurl: Human-induced Curriculum Discovery (2023.acl-long)

Copied to clipboard

Challenge: Existing curriculum learning frameworks can be used to discover effective curricula for NLP tasks based on prior knowledge about sample difficulty.
Approach: They propose a framework for curriculum learning based on prior knowledge about sample difficulty.
Outcome: The proposed framework outperforms existing curriculum learning approaches on several NLP tasks and can prune and weight samples for better learning.
Efficient Contrastive Learning via Novel Data Augmentation and Curriculum Learning (2021.emnlp-main)

Copied to clipboard

Challenge: Recent studies describe how to apply contrastive learning to the language domain but it is difficult to apply data augmentation methods directly to language modeling.
Approach: They propose a memory-efficient continual pretraining method that applies contrastive learning with novel data augmentation and curriculum learning.
Outcome: The proposed method outperforms baseline models on sentence-level tasks with only 70% of memory compared to the baseline model.
EPiDA: An Easy Plug-in Data Augmentation Framework for High Performance Text Classification (2022.naacl-main)

Copied to clipboard

Challenge: Existing methods for data augmentation do not fully exploit the potential of DA in NLP.
Approach: They propose an easy and plug-in framework for data augmentation to support effective text classification.
Outcome: The proposed framework outperforms existing methods in most cases, but not using agent networks or pre-trained generation networks.
Dynamic Augmentation Data Selection for Few-shot Text Classification (2022.findings-emnlp)

Copied to clipboard

Challenge: Data augmentation is a popular method for fine-tuning pre-trained language models to increase model robustness and performance.
Approach: They propose a dynamic data selection method to select effective augmentation data from different augmentation sources according to the model’s learning stage by identifying a set of augmentation samples that optimally facilitates the learning process of the most current model.
Outcome: The proposed method outperforms strong baselines on a variety of sentence classification tasks.
Reinforcement Learning based Curriculum Optimization for Neural Machine Translation (N19-1)

Copied to clipboard

Challenge: a heterogeneous training dataset can vary in characteristics such as domain, translation quality, and degree of difficulty.
Approach: They propose to use reinforcement learning to learn an optimal curriculum for NMT training . they find it can beat uniform baselines and hand-designed, state-of-the-art curricula .
Outcome: The proposed approach beats baselines and hand-designed curricula on English-to-French datasets.
Dynamic Data Selection and Weighting for Iterative Back-Translation (2020.emnlp-main)

Copied to clipboard

Challenge: Experimental results demonstrate that our methods achieve improvements of up to 1.8 BLEU points over competitive baselines.
Approach: They propose a data selection and weighting strategy to iterate back-translation models and apply it to it . they use a target language to back-transcribe monolingual data, which is of high quality and reflect the target domain.
Outcome: The proposed approach achieves 1.8 BLEU points over baselines on domain adaptation, low-resource, and high-resourced MT settings and on two language pairs.
A Data Cartography based MixUp for Pre-trained Language Models (2022.naacl-main)

Copied to clipboard

Challenge: Using MixUp, additional samples are generated during training by combining random pairs of training samples and their labels.
Approach: They propose a new MixUp strategy that leverages Training Dynamics and allows more informative samples to be combined for generating new data samples.
Outcome: The proposed method achieves competitive performance using a smaller subset of training data compared with strong baselines and yields lower expected calibration error on the pre-trained language model, BERT, on both in-domain and out-of-domain settings.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations