Improving Scheduled Sampling with Elastic Weight Consolidation for Neural Machine Translation (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Autoregressive models trained with maximum likelihood estimation suffer from exposure bias, i.e. the discrepancy between ground-truth prefixes used during training and model-generated prefix at inference time. |
| Approach: | They propose to use Elastic Weight Consolidation to better balance mitigating exposure bias with retaining performance. |
| Outcome: | The proposed method significantly outperforms maximum likelihood estimation and scheduled sampling baselines on four translation datasets. |
Similar Papers
Confidence-Aware Scheduled Sampling for Neural Machine Translation (2021.findings-acl)
Copied to clipboard
| Challenge: | Neural machine translation models are trained to maximize the likelihood of the next token given previous golden tokens as inputs, but at the inference stage, golden token is unavailable. |
| Approach: | They propose a scheduled sampling method that randomly replaces groundtruth tokens with predicted ones during training, ignoring real-time model competence. |
| Outcome: | The proposed method outperforms the Transformer and vanilla scheduled sampling on large-scale translations. |
Scheduled Sampling Based on Decoding Steps for Neural Machine Translation (2021.emnlp-main)
Copied to clipboard
| Challenge: | Neural machine translation models are trained to maximize the likelihood of next token given previous golden tokens as inputs, but at the inference stage, golden token is unavailable. |
| Approach: | They propose to use scheduled sampling to replace ground-truth tokens with predicted tokens to bridge the gap between training and inference. |
| Outcome: | The proposed methods outperform the Transformer baseline and vanilla scheduled sampling on three large-scale WMT tasks. |
Scheduled Sampling for Transformers (P19-2)
Copied to clipboard
| Challenge: | Existing studies show that scheduled sampling can be applied to recurrent neural networks to avoid exposure bias. |
| Approach: | They propose to use teacher forced embeddings and model predictions to avoid exposure bias in sequence-to-sequence generation. |
| Outcome: | The proposed technique achieves performance close to a teacher-forcing baseline on two language pairs and is promising for future research. |
Differentiable Sampling with Flexible Reference Word Order for Neural Machine Translation (N19-1)
Copied to clipboard
| Challenge: | Existing approaches to correct exposure bias in machine translation are inadequate . scheduled sampling assumes that words are aligned at each time step . |
| Approach: | They propose a differentiable sampling algorithm that optimizes the probability that the reference can be aligned with the sampled output. |
| Outcome: | The proposed approach improves BLEU on translation tasks and is simpler to train with no sampling schedule. |
Mitigating the Diminishing Effect of Elastic Weight Consolidation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing work addresses catastrophic forgetting in sequential training by fine-tuning pre-trained language models on different datasets. |
| Approach: | They propose to rescale the components of EWC to mitigate catastrophic forgetting by mixing new and old training data and retraining the model from scratch. |
| Outcome: | The proposed method requires smaller values for the trade-off parameters to achieve comparable results to EWC on natural language inference and fact-checking tasks. |
Overcoming Catastrophic Forgetting During Domain Adaptation of Neural Machine Translation (N19-1)
Copied to clipboard
| Challenge: | Neural Machine Translation (NMT) performs poorly without large training corpora. |
| Approach: | They propose a machine learning method that retains the majority of general-domain performance lost in continued training without degrading in-domain. |
| Outcome: | The proposed method retains the majority of general-domain performance lost in continued training without degrading in-domain performances. |
An Efficient Rehearsal Scheme for Catastrophic Forgetting Mitigation during Multi-stage Fine-tuning (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing approaches to fine-tune foundational models on new tasks or domains are costly and time-consuming. |
| Approach: | They propose a sampling scheme that prioritizes rehearsal of "collateral damage" samples . the scheme is computationally efficient and easy to implement, they say . |
| Outcome: | a new approach prioritizes rehearsal of “collateral damage” samples outperforms other continual learning methods. |
Alleviating Exposure Bias in Abstractive Summarization via Sequentially Generating and Revising (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing approaches to abstractive summarization suffer from exposure bias . Existing solutions bridge this gap through un- or semi-supervised holistic learning . |
| Approach: | They propose to reformat abstractive summarization to sequential generation and revision (SeGRe) this allows the model to assess the flawed summary from a global perspective and modify inappropriate expressions. |
| Outcome: | The proposed model can assess the flawed summary from a global view and modify inappropriate expressions. |
Mitigating Exposure Bias in Grammatical Error Correction with Data Augmentation and Reweighting (2023.eacl-main)
Copied to clipboard
| Challenge: | Existing approaches to grammatical error correction (GEC) use sequence-to-sequence models, but there is an exposure bias problem. |
| Approach: | They propose a data manipulation approach to overcome the exposure bias problem in seq2seq GEC . they propose augmentation methods to mimic decoder input and reweighting methods to automatically balance the importance of each kind of augmented samples. |
| Outcome: | The proposed method improves on benchmark GEC datasets. |
Bridging the Gap between Training and Inference for Neural Machine Translation (P19-1)
Copied to clipboard
| Challenge: | Neural Machine Translation generates target words sequentially while at inference it has to generate the entire sequence from scratch. |
| Approach: | They propose to use ground truth and inference to generate target words sequentially while at inference it has to generate the entire sequence from scratch. |
| Outcome: | Experiments on Chinese->English and WMT’14 English->German translation tasks show that the proposed model can achieve significant improvements on multiple datasets. |