Challenge: Maximum likelihood estimation (MLE) is used to train models, but during testing, the model is conditioned on previously generated tokens, resulting in exposure bias.
Approach: They propose to use optimal transport to match the sequences generated in MLE and test modes to reduce exposure bias.
Outcome: The proposed method is validated on machine translation, text summarization, and text generation tasks.

Similar Papers

Best Student Forcing: A Simple Training Mechanism in Adversarial Language Generation (2020.lrec-1)

Copied to clipboard

Challenge: Language models trained with Maximum Likelihood Estimation (MLE) have been considered as a mainstream solution in Natural Language Generation (NLG) however, they are reportedly suffering from training instability and mode collapse, and therefore outperform conventional MLE models.
Approach: They propose a method to improve Generative Adversarial Nets (GANs) using best student forcing and discriminators to increase training stability and sample diversity.
Outcome: The proposed techniques outperform MLE models and outperformed existing approaches in terms of sample diversity and training stability.
Implicit Unlikelihood Training: Improving Neural Text Generation with Reinforcement Learning (2021.eacl-main)

Copied to clipboard

Challenge: Existing approaches to language modeling use autoregressive methods, but they can produce repetitive results.
Approach: They propose to add a loss function for regularization to avoid unwanted properties, such as contradiction or repetition, to a language model by using policy gradient reinforcement learning.
Outcome: The proposed method reduces repetition without impacting the language model quality.
Addressing the Training-Inference Discrepancy in Discrete Diffusion for Text Generation (2025.coling-main)

Copied to clipboard

Challenge: Existing discrete diffusion models for text generation have a discrepancy between training and inference.
Approach: They propose a training schema that considers two-step diffusion processes and a scheduling technique that gradually increases the probability of using self-generated text as training progresses.
Outcome: The proposed training schema and scheduling technique improve diffusion models on four widely used datasets.
Plug-in Language Model: Controlling Text Generation with a Simple Regression Model (2024.findings-naacl)

Copied to clipboard

Challenge: Large-scale pre-trained language models have demonstrated unrivaled capacity in generating text that closely resembles human-written content.
Approach: They propose a plug-in language model that leverages reinforcement learning to adjust latent states to control text generation.
Outcome: The proposed model outperforms existing methods that rely on gradient-based, weighted decoding, or prompt-based methods.
End-to-End Bias Mitigation by Modelling Biases in Corpora (2020.acl-main)

Copied to clipboard

Challenge: Recent studies have shown that strong natural language understanding models are prone to relying on unwanted dataset biases without learning the underlying task.
Approach: They propose two learning strategies to train neural models that are more robust to dataset biases and transfer better to out-of-domain datasets.
Outcome: The proposed methods improve robustness in all settings and transfer better to out-of-domain datasets.
Click: Controllable Text Generation with Sequence Likelihood Contrastive Learning (2023.findings-acl)

Copied to clipboard

Challenge: Current language models have shown impressive capability of generating fluent and grammatical text, but they often produce behaviors misaligned with human expectations.
Approach: They propose a new language model called Leo for controllable text generation which employs a contrastive loss on sequence likelihood which fundamentally decreases the generation probability of negative samples.
Outcome: The proposed model outperforms baselines on language detoxification, sentiment steering, and repetition reduction tasks.
Overestimation of Syntactic Representation in Neural Language Models (2020.acl-main)

Copied to clipboard

Challenge: Several testing methodologies have been developed to probe models’ syntactic representations.
Approach: They propose a method to determine syntactic structure by training a model on strings generated according to a template and testing its ability to distinguish between similar ones with different syntax.
Outcome: The proposed method reproduces positive results with two non-syntactic baseline language models: an n-gram model and an LSTM model trained on scrambled inputs.
Generating Temporally-ordered Event Sequences via Event Optimal Transport (2022.coling-1)

Copied to clipboard

Challenge: Existing methods for temporal event ordering and event infilling ignore the global semantics of events, and the model adopts a word-level objective to model events in texts.
Approach: They propose a temporal event ordering and event infilling task using a model that uses maximum likelihood estimation to model events in texts.
Outcome: The proposed model outperforms existing models on all evaluation datasets.
Neural Mask Generator: Learning to Generate Adaptive Word Maskings for Language Model Adaptation (2020.emnlp-main)

Copied to clipboard

Challenge: Existing methods to train language models on diverse text corpora have brought up performance improvements on several natural language understanding (NLU) tasks.
Approach: They propose a method to automatically generate domain- and task-adaptive maskings of a given text for self-supervised pre-training.
Outcome: The proposed framework outperforms rule-based masking strategies on question answering and text classification datasets on which it outperformed rule-driven masking techniques.
Leveraging Training Dynamics and Self-Training for Text Classification (2022.findings-emnlp)

Copied to clipboard

Challenge: Semi-supervised learning (SSL) is a promising technique for improving deep learning models when training data is scarce.
Approach: They propose a semi-supervised learning approach that leverages training dynamics of unlabeled data.
Outcome: The proposed method achieves an average increase in F1 score of 3.5% over baselines in low resource 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