The recently proposed mask-predict decoding algorithm has narrowed the performance gap between semi-autoregressive machine translation models and the traditional left-to-right approach. We introduce a new training method for conditional masked language models, SMART, which mimics the semi-autoregressive behavior of mas…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,695 papers · 148 categories
Trend · papers per month
48 results for “autoregressive training”
Paper proposes AXE loss for non-autoregressive machine translation, improving performance.
problem Challenges in training non-autoregressive models due to lack of autoregressive factors and cross entropy loss penalties.
method Proposes aligned cross entropy (AXE) loss function using a differentiable dynamic program for better word order alignment.
result AXE-based training improves performance on major WMT benchmarks and sets a new state of the art for non-autoregressive models.
Paper proposes energy objective for training normalizing flows without determinants.
problem Challenges in training normalizing flows due to Jacobian determinants.
method Introduces energy objective based on proper scoring rules, determinant-free.
result Energy objective supports novel model families and competitive performance.
Standard autoregressive seq2seq models are easily trained by max-likelihood, but tend to show poor results under small-data conditions. We introduce a class of seq2seq models, GAMs (Global Autoregressive Models), which combine an autoregressive component with a log-linear component, allowing the use of global \textit{a…
Neural autoregressive models are explicit density estimators that achieve state-of-the-art likelihoods for generative modeling. The D-dimensional data distribution is factorized into an autoregressive product of one-dimensional conditional distributions according to the chain rule. Data completion is a more involved ta…
Parallelizes autoregressive generation using VSSM.
problem Autoregressive models' inability to parallelize generation.
method Variational SSM (VSSM) with parallelizable sampling and decoding.
result Parallel generation possible with VSSM.
Transformers learn a mesa-optimizer to implement in-context learning.
problem Understanding the convergence of autoregressive training to a mesa-optimizer.
method Investigated a one-layer linear causal self-attention model autoregressively trained by gradient flow.
result Proved that autoregressive training converges to a gradient descent step for an OLS problem, validating the mesa-optimizer hypothesis.
Autoregressive state transitions, where predictions are conditioned on past predictions, are the predominant choice for both deterministic and stochastic sequential models. However, autoregressive feedback exposes the evolution of the hidden state trajectory to potential biases from well-known train-test discrepancies.…
Due to the unparallelizable nature of the autoregressive factorization, AutoRegressive Translation (ART) models have to generate tokens sequentially during decoding and thus suffer from high inference latency. Non-AutoRegressive Translation (NART) models were proposed to reduce the inference time, but could only achiev…
Non-autoregressive translation (NAT) models remove the dependence on previous target tokens and generate all target tokens in parallel, resulting in significant inference speedup but at the cost of inferior translation accuracy compared to autoregressive translation (AT) models. Considering that AT models have higher a…
This work proposes an efficient autoregressive model for text generation.
problem The challenge of generating high-quality text with autoregressive models.
method Introduces a cascaded decoding approach using Markov transformers to achieve sub-linear parallel time generation.
result Shows competitive accuracy/speed tradeoff compared to existing methods on five machine translation datasets.
Autoregressive models struggle with hard-to-compute distributions, alternatives like energy-based and latent-variable models solve this.
problem Autoregressive models struggle with distributions whose next-symbol probability is hard to compute.
method Alternatives include energy-based models and latent-variable autoregressive models.
result Alternatives to autoregressive models can escape limitations of hard-to-compute distributions.
Autoregressive generative models of images tend to be biased towards capturing local structure, and as a result they often produce samples which are lacking in terms of large-scale coherence. To address this, we propose two methods to learn discrete representations of images which abstract away local detail. We show th…
New model handles missing data effectively in autoregressive models.
problem Handling missing data in autoregressive models.
method Reinterpret existing models through missing data lens, introduce principled framework for incomplete datasets, active information acquisition.
result MO-ARM consistently outperforms imputation baselines across real-world benchmarks.
We present the Insertion Transformer, an iterative, partially autoregressive model for sequence generation based on insertion operations. Unlike typical autoregressive models which rely on a fixed, often left-to-right ordering of the output, our approach accommodates arbitrary orderings by allowing for tokens to be ins…
Autoregressive networks can achieve promising performance in many sequence modeling tasks with short-range dependence. However, when handling high-dimensional inputs and outputs, the huge amount of parameters in the network lead to expensive computational cost and low learning efficiency. The problem can be alleviated …
This paper explores how Transformers predict next tokens in autoregressive tasks.
problem Understanding the success of Transformers in autoregressive learning.
method Trained a Transformer on a next-token prediction task, focusing on commuting orthogonal matrices.
result Trained Transformers can be seen as implementing gradient descent for a specific objective function.
ARCNPs improve CNPs by autoregressively modeling dependencies.
problem CNPs struggle with modeling dependencies in predictions.
method Autoregressive deployment of factorized Gaussian CNPs.
result ARCNPs significantly outperform non-AR CNPs in various tasks.
ARDMs are a new model class for autoregressive diffusion that generalize existing models and can compress data efficiently.
problem Efficient data compression and generation.
method Autoregressive Diffusion Models (ARDMs) that generalize existing autoregressive models and discrete diffusion models.
result ARDMs require significantly fewer steps for compression compared to discrete diffusion models.
Normalizing flows and autoregressive models have been successfully combined to produce state-of-the-art results in density estimation, via Masked Autoregressive Flows (MAF), and to accelerate state-of-the-art WaveNet-based speech synthesis to 20x faster than real-time, via Inverse Autoregressive Flows (IAF). We unify a…
Diffusion models generate music sequences without autoregressive loops.
problem Generating music sequences from symbolic data using diffusion models.
method Parameterize discrete symbolic data in continuous latent space, train diffusion model, generate sequences through reverse process.
result Strong unconditional generation and post-hoc conditional infilling compared to autoregressive models.
AR-CSM models use derivatives of univariate log-conditionals to estimate joint distributions efficiently.
problem Scalability and stability issues in training autoregressive models.
method Parameterize joint distribution using derivatives of univariate log-conditionals and introduce Composite Score Matching (CSM) for efficient training.
result AR-CSM models are more scalable and stable compared to previous score matching algorithms.
Missing value imputation is a fundamental problem in spatiotemporal modeling, from motion tracking to the dynamics of physical systems. Deep autoregressive models suffer from error propagation which becomes catastrophic for imputing long-range sequences. In this paper, we take a non-autoregressive approach and propose …
This review compares various deep generative models.
problem Training deep neural networks to model data distributions.
method Comprehensive comparison of VAEs, GANs, flows, energy models, and autoregressives.
result Trade-offs and interrelationships among different models.
Improved simulation of phase transitions using hierarchical autoregressive networks.
problem Simulating phase transitions in complex systems.
method Hierarchical Autoregressive Neural (HAN) network sampling algorithm.
result Significant improvement in statistical uncertainty compared to the Wolff cluster algorithm.
Efficiently combines autoregressive and set-based models for joint distributions.
problem Joint distributions over multiple predictions from set-based models.
method Causal autoregressive buffer that caches context and captures dependencies.
result Up to 20x faster joint sampling and density evaluation, up to 7x lower memory usage.
New approach uses autoregressive models to explore and quantify uncertainty in decision-making.
problem Quantifying and exploring uncertainty in online decision-making.
method Reformulates uncertainty as missing future outcomes, training autoregressive models for next-outcome prediction.
result Establishes a reduction from online learning to offline next-outcome prediction, controlling Bayesian regret by sequence prediction loss.
Improved lattice field theory simulations with local-Autoregressive Conditional Normalizing Flow.
problem Efficiently sampling lattice field theories with computational challenges.
method Integrates locality into autoregressive conditional normalizing flows.
result Autocorrelation times improved by orders of magnitude for theory on a 2D lattice.
Modified asymmetric hidden Markov models for time series with autoregressive components.
problem Dynamic relationships between variables in time series data.
method Introducing an asymmetric autoregressive component to recent asymmetric hidden Markov models.
result The model can choose the optimal autoregressive order for better likelihood.
Efficient autoregressive entity linking with correction for faster, more accurate results.
problem High computational cost and non-parallelizable decoding in autoregressive entity linking.
method Parallelizes autoregressive linking across all mentions, uses a shallow decoder, and adds a discriminative correction term.
result 70 times faster and more accurate than previous methods, outperforming state-of-the-art approaches.
Neural density estimators are flexible families of parametric models which have seen widespread use in unsupervised machine learning in recent years. Maximum-likelihood training typically dictates that these models be constrained to specify an explicit density. However, this limitation can be overcome by instead using …
This study compares different types of normalizing flows for generating complex distributions.
problem Comparing different types of normalizing flows for generating complex distributions.
method Real-valued non-Volume preserving (RealNVP), masked autoregressive flow (MAF), coupling rational quadratic spline (C-RQS), and autoregressive rational quadratic spline (A-RQS) were compared using statistical tests.
result A-RQS algorithm outperforms others in terms of accuracy and training speed.
A new autoregressive model learns the order of graph generation tasks.
problem Generating graphs in a meaningful order when the canonical order is not obvious.
method Introduces a variant of autoregressive models that dynamically decides the autoregressive order based on data.
result Achieves state-of-the-art results on molecular graph generation benchmarks.
Convolutional autoregressive models have recently demonstrated state-of-the-art performance on a number of generation tasks. While fast, parallel training methods have been crucial for their success, generation is typically implemented in a naïve fashion where redundant computations are unnecessarily repeated. This res…
Study finds optimal vocabulary size for neural machine translation.
problem Imbalanced class distribution in language data affects NMT performance.
method Casts NMT as a classification task, analyzes vocabulary sizes, and tests multiple languages.
result Certain vocabulary sizes outperform others, explaining NMT performance.
Improved autoregressive models generate higher quality images and are more robust to noise.
problem Generating high-quality images from autoregressive models.
method Noise conditional maximum likelihood estimation (MLE) with score-based sampling.
result Models trained with noise conditional MLE achieve better test likelihoods and generate higher quality images.
mGENRE improves multilingual entity linking with autoregressive sequence prediction.
problem Multilingual Entity Linking (MEL) task of resolving language-specific mentions to a multilingual Knowledge Base.
method Autoregressive sequence-to-sequence system that cross-encodes mention strings and entity names.
result Over 50% improvement in average accuracy in zero-shot settings.
Study evaluates stock price forecasting models during the pandemic.
problem Forecasting stock prices during the Covid-19 pandemic.
method Four models (Long-Short Term Memory, XGBoost, Autoregression, Last Value) were tested on stock prices of Facebook, Amazon, Tesla, Google, and Apple.
result Autoregression and Last Value models outperform other models due to strong correlation between prices.
WAVE improves time series forecasting by integrating AR and MA components.
problem Time series forecasting challenges.
method WAVE attention mechanism with AR and MA components.
result WAVE attention consistently improves TSF performance.
Efficient methods for answering complex probabilistic queries in sequential data.
problem Complex probabilistic queries in sequential data.
method Broad class of novel approximation techniques for marginalization in sequential models.
result Efficient techniques for answering long-range probabilistic queries.
Autoregressive flow models can perform causal discovery and inference tasks.
problem Causal inference tasks such as causal discovery and interventional predictions.
method Using autoregressive flow models to estimate causal directions and make predictions.
result Autoregressive flows can accurately perform causal inference tasks without restrictive assumptions.
We present Sequential Neural Likelihood (SNL), a new method for Bayesian inference in simulator models, where the likelihood is intractable but simulating data from the model is possible. SNL trains an autoregressive flow on simulated data in order to learn a model of the likelihood in the region of high posterior dens…
Identification of a groundwater contaminant source simultaneously with the hydraulic conductivity in highly-heterogeneous media often results in a high-dimensional inverse problem. In this study, a deep autoregressive neural network-based surrogate method is developed for the forward model to allow us to solve efficien…
BiGG model efficiently generates sparse graphs with reduced complexity.
problem Challenges in scalable deep learning for sparse graphs.
method BiGG model, an autoregressive model that leverages graph sparsity.
result Graph generation time complexity reduced from to .
GAR generalizes autoregression for efficient multi-fidelity fusion.
problem Efficiently combining low-fidelity and high-fidelity simulation results.
method Generalized autoregression (GAR) using tensor formulation and latent features.
result GAR outperforms state-of-the-art methods with a large margin in RMSE.
LTMs use latent vectors for efficient autoregressive generation.
problem Efficient autoregressive generation in language models.
method Dual-rate optimization in variational Bayes framework.
result LTMs achieve superior sample and parameter efficiency.
Listen and Fill in the Missing Letters: Non-Autoregressive Transformer for Speech Recognitioneess.AS
Recently very deep transformers have outperformed conventional bi-directional long short-term memory networks by a large margin in speech recognition. However, to put it into production usage, inference computation cost is still a serious concern in real scenarios. In this paper, we study two different non-autoregressi…
What Language Model Architecture and Pretraining Objective Work Best for Zero-Shot Generalization?cs.CL
New study finds best language model architecture and pretraining objective for zero-shot tasks.
problem Evaluating which language model architectures and pretraining objectives best enable zero-shot generalization.
method Compared three model architectures and two pretraining objectives across 170 billion tokens, with and without finetuning.
result Causal decoder-only models trained on autoregressive language modeling exhibit strongest zero-shot generalization.