AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Learning to Scale Multilingual Representations for Vision-Language Tasks

Andrea Burns, Donghyun Kim, Derry Wijaya, Kate Saenko, Bryan A. Plummer

ECCV 2020 · Vision-language and multimodal › Human and embodied modeling · approach: Vision-language pretraining · 34 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
add language-specific embeddings to the language-agnostic vocabulary (hybrid embedding model) (ablation)Language-Agnostic · MSCOCO multilingual retrievalaverage mR (A): 73.5 -> 76.9Improved
add language-specific embeddings to the language-agnostic vocabulary (hybrid embedding model) (ablation)Language-Agnostic · Multi30K multilingual retrievalaverage mR (A): 61.6 -> 64.0Improved
add Masked Cross-Language Modeling loss (ablation)HEM · MSCOCO multilingual retrievalaverage mR (A): 76.9 -> 77.7Improved
add Masked Cross-Language Modeling loss (ablation)HEM · Multi30K multilingual retrievalaverage mR (A): 64.0 -> 65.4Improved
add CLC-A (average ensemble over translated queries) (ablation)SMALR · MSCOCO multilingual retrievalhuman-average mR (HA): 77.7 -> 78.6Improved
add CLC-A (average ensemble over translated queries) (ablation)SMALR · Multi30K multilingual retrievalhuman-average mR (HA): 65.0 -> 68.2Improved
replace CLC-A with weighted CLC-C classifier ensemble (ablation)SMALR-CLC-A · MSCOCO multilingual retrievalhuman-average mR (HA): 78.6 -> 79.7Improved
replace CLC-A with weighted CLC-C classifier ensemble (ablation)SMALR-CLC-A · Multi30K multilingual retrievalhuman-average mR (HA): 68.2 -> 68.7Improved
add exploration parameters (p=0.2, M=20) when mapping to the latent vocabulary (ablation)Language-Agnostic · Multi30K multilingual retrievalaverage mR (A): 72.1 -> 73.5Improved
add exploration parameters (p=0.2, M=20) when mapping to the latent vocabulary (ablation)Language-Agnostic · MSCOCO multilingual retrievalaverage mR (A): 60.0 -> 61.6Improved
vary the weight lambda_2 of the masking loss (ablation)SMALR · Multi30K multilingual retrievalaverage mRNo clear effect
use SMALR instead of S-LIWE as the multilingual representationS-LIWE (10) · MSCOCO multilingual retrievalaverage mR (A): 66.7 -> 77.7Improved

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