AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Discovering Non-monotonic Autoregressive Orderings with Variational Inference

Xuanlin Li, Brandon Trabucco, Dong Huk Park, Michael Luo, Sheng Shen, Trevor Darrell, Yang Gao

ICLR 2021 · Probabilistic and scientific modeling › Molecules and proteins · approach: Neuro-symbolic probabilistic layers · 1 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
replace fixed generation order with learned VOI orderTransformer-InDIGO on MS-COCO · MS-COCOCIDEr: 95.6 -> 100.6Improved
replace fixed generation order with learned VOI orderTransformer-InDIGO on Django · DjangoBLEU: 42.6 -> 45.9Improved
replace fixed generation order with learned VOI orderTransformer-InDIGO on Gigaword · GigawordR-1: 35.6 -> 36.6Improved
replace L2R order with learned VOI orderTransformer-InDIGO on WMT16 Ro-En · WMT16 Ro-EnBLEU: 32.7 -> 32.9Improved
use VOI instead of the strongest reported baselineTransformer-InDIGO on WMT16 Ro-En · WMT16 Ro-EnTER: 49.0 -> 49.3Hurt
use Gumbel-Matching instead of Plackett-Luce (ablation)VOI encoder order recovery · MS-COCO order recoveryNormalized Levenshtein Distance: 0.62 -> 0.25Improved
use Transformer-XL relative positional encoding instead of sinusoid encoding (ablation)VOI encoder with Gumbel-Matching · MS-COCO order recoveryNormalized Levenshtein Distance: 0.40 -> 0.25Improved
increase K from 2 to 10 sampled latents (ablation)VOI encoder order recovery · MS-COCO order recoveryNormalized Levenshtein Distance: 0.31 -> 0.21Improved
increase K from 10 to 20 sampled latents (ablation)VOI encoder order recovery · MS-COCO order recoveryNormalized Levenshtein Distance: 0.21 -> 0.21No clear effect
fine-tune the decoder with the encoder fixed (ablation)VOI on MS-COCO · MS-COCOperformanceHurt
fine-tune the decoder with the encoder fixed (ablation)VOI on Gigaword · GigawordROUGE: around 1.5 to 2.0Improved
fine-tune the decoder with the encoder fixed (ablation)VOI on WMT16 Ro-En · WMT16 Ro-EnBLEU: around 1.5 pointsImproved

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