Fast Decoding in Sequence Models Using Discrete Latent Variables
E899037
"Fast Decoding in Sequence Models Using Discrete Latent Variables" is a research paper that introduces a method for accelerating sequence model inference by leveraging discrete latent representations to enable more parallelizable decoding.
All labels observed (1)
| Label | Occurrences |
|---|---|
| Fast Decoding in Sequence Models Using Discrete Latent Variables canonical | 1 |
How this entity was disambiguated
This entity first appeared as the object of triple T11003401 — resolving that mention is where its identity was fixed. The disambiguator weighed these candidate entities and picked the highlighted one (or “None”, minting a new entity). This is how homonymy is resolved: the same surface form can point to different entities.
NED1
Entity disambiguation (via context triple)
gpt-5-mini-2025-08-07
Target entity: Fast Decoding in Sequence Models Using Discrete Latent Variables Context triple: [Łukasz Kaiser, coAuthorOf, Fast Decoding in Sequence Models Using Discrete Latent Variables]
-
A.
Neural Discrete Representation Learning
Neural Discrete Representation Learning is a machine learning framework that introduces Vector Quantized Variational Autoencoders (VQ-VAE) to learn discrete latent representations for high-dimensional data such as images, audio, and video.
-
B.
Exploring the Limits of Language Modeling
"Exploring the Limits of Language Modeling" is a research paper that investigates how far large-scale neural language models can be pushed in terms of performance, scalability, and generalization on natural language tasks.
-
C.
Auto-Encoding Variational Bayes
Auto-Encoding Variational Bayes is the foundational 2013 paper by Kingma and Welling that introduced variational autoencoders, a generative model framework combining deep learning with variational Bayesian inference.
-
D.
Sequence to Sequence Learning with Neural Networks
"Sequence to Sequence Learning with Neural Networks" is a seminal 2014 paper that introduced the sequence-to-sequence (seq2seq) neural network framework for tasks like machine translation, laying the groundwork for many modern NLP models.
-
E.
VQ-VAE
VQ-VAE is a neural network model that combines vector quantization with variational autoencoders to learn discrete latent representations for tasks like image and audio generation.
- F. None of above. chosen
- G. Unsure - the case is ambiguous/there is not enough information to decide.
NED2
Entity disambiguation (via description)
gpt-5-mini-2025-08-07
Target entity: Fast Decoding in Sequence Models Using Discrete Latent Variables Target entity description: "Fast Decoding in Sequence Models Using Discrete Latent Variables" is a research paper that introduces a method for accelerating sequence model inference by leveraging discrete latent representations to enable more parallelizable decoding.
-
A.
Neural Discrete Representation Learning
Neural Discrete Representation Learning is a machine learning framework that introduces Vector Quantized Variational Autoencoders (VQ-VAE) to learn discrete latent representations for high-dimensional data such as images, audio, and video.
-
B.
Exploring the Limits of Language Modeling
"Exploring the Limits of Language Modeling" is a research paper that investigates how far large-scale neural language models can be pushed in terms of performance, scalability, and generalization on natural language tasks.
-
C.
Auto-Encoding Variational Bayes
Auto-Encoding Variational Bayes is the foundational 2013 paper by Kingma and Welling that introduced variational autoencoders, a generative model framework combining deep learning with variational Bayesian inference.
-
D.
Sequence to Sequence Learning with Neural Networks
"Sequence to Sequence Learning with Neural Networks" is a seminal 2014 paper that introduced the sequence-to-sequence (seq2seq) neural network framework for tasks like machine translation, laying the groundwork for many modern NLP models.
-
E.
VQ-VAE
VQ-VAE is a neural network model that combines vector quantization with variational autoencoders to learn discrete latent representations for tasks like image and audio generation.
- F. None of above. chosen
Statements (30)
| Predicate | Object |
|---|---|
| instanceOf |
research paper
ⓘ
scientific publication ⓘ |
| addresses |
computational cost of sequential decoding
ⓘ
latency in sequence model inference ⓘ |
| aimsTo |
improve inference efficiency
ⓘ
reduce decoding time ⓘ |
| appliesTo |
autoregressive sequence models
ⓘ
neural sequence-to-sequence models ⓘ |
| assumes | availability of a learned discrete latent space ⓘ |
| comparesWith | standard autoregressive sequence models ⓘ |
| contribution | introduces discrete latent structure to enable more parallel decoding ⓘ |
| evaluatedIn | sequence generation tasks ⓘ |
| field |
machine learning
ⓘ
natural language processing ⓘ |
| focusesOn |
discrete latent variables
ⓘ
fast decoding ⓘ parallelizable decoding ⓘ sequence models ⓘ |
| improves | decoding speed compared to standard autoregressive decoding ⓘ |
| motivatedBy | need for low-latency sequence generation ⓘ |
| proposes | method for accelerating sequence model inference ⓘ |
| relatedTo |
discrete representation learning
ⓘ
language modeling ⓘ latent variable models ⓘ neural machine translation ⓘ parallel decoding strategies ⓘ |
| targets | faster inference at test time ⓘ |
| typeOfDecoding | non-strictly-autoregressive decoding ⓘ |
| uses |
discrete latent representations
ⓘ
neural networks ⓘ |
How these facts were elicited
The pipeline generated the facts above by prompting gpt-5.1 with this entity's name + description and the instruction below.
Instruction
You are a knowledge base construction expert. Given a subject entity and a description of it, return factual statements that you know for the subject as a JSON list of dictionaries(triples), where keys must be "subject", "predicate" and "object". The number of facts may be very high, between 25 to 50 or more, for very popular subjects. For less popular subjects, the number of facts can be very low, like 5 or 10. # Requirements - If you don't know the subject at all, return an empty list. - If the subject is not a named entity, return an empty list. - Include at least one triple where predicate is "instanceOf". - Do not get too wordy. - Separate several objects into multiple triples with one object.
Input
Subject: Fast Decoding in Sequence Models Using Discrete Latent Variables Description of subject: "Fast Decoding in Sequence Models Using Discrete Latent Variables" is a research paper that introduces a method for accelerating sequence model inference by leveraging discrete latent representations to enable more parallelizable decoding.
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.
subject linked to:
Lukasz Kaiser