AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Sequence Modeling with Unconstrained Generation Order

Dmitrii Emelianenko, Elena Voita, Pavel Serdyukov

NeurIPS 2019 · Image generation › Autoregressive and latent generation · approach: Generative latent space tooling · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use unconstrained insertion order instead of fixed-order decodingTransformer-base machine translation · WMT2017 En-RuBLEU: 31.6 -> 33.2Improved
Use unconstrained insertion order instead of fixed-order decodingTransformer-base machine translation · WMT2017 Ru-EnBLEU: 35.3 -> 36.4Improved
Use unconstrained insertion order instead of the strongest fixed orderTransformer-base machine translation · ASPEC En-Ja testBLEU: 48.6 -> 50.3Improved
Use unconstrained insertion order instead of fixed-order decodingTransformer-base machine translation · IWSLT En-ArBLEU: 12.0 -> 12.2No clear effect
Use unconstrained insertion order instead of fixed-order decodingTransformer-base machine translation · IWSLT En-DeBLEU: 28.04 -> 28.36Improved
Use unconstrained insertion order instead of fixed-order decodingTransformer-base machine translation · IWSLT De-EnBLEU: 33.17 -> 33.08No clear effect
Use unconstrained insertion order instead of fixed-order decodingTransformer-base Image-to-LaTeX · ImageToLatex-140KBLEU: 89.5 -> 90.3Improved
Use unconstrained insertion order instead of fixed-order decodingTransformer-base image captioning · MSCOCOBLEU: 18.0 -> 25.6Improved
Use unconstrained insertion order instead of fixed-order decodingTransformer-base image captioning · MSCOCOCIDEr: 56.1 -> 81.0Improved
Add model-dependent trajectory sampling training after pretraining (ablation)INTRUS · WMT En-Ru 2012-2013 testBLEU: 24.6 -> 27.5Improved
Use sampled trajectory training instead of beam-search argmax trajectory training (ablation)INTRUS · WMT En-Ru 2012-2013 testBLEU: 26.6 -> 27.5Improved
Use uniform feasible-trajectory pretraining rather than left-to-right pretraining (ablation)INTRUS · WMT En-Ru 2012-2013 testBLEU: 26.3 -> 27.5Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy