AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Large-scale Pretraining for Visual Dialog: A Simple State-of-the-Art Baseline

Vishvak Murahari, Dhruv Batra, Devi Parikh, Abhishek Das

ECCV 2020 · Vision-language and multimodal › Visual question answering · approach: Vision-language model VQA evaluation · 110 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 MLM loss to NSP pretraining (ablation)language-only BERT · VisDial v1.0 valNDCG: 56.17 -> 57.26Improved
increase dialog history from 0 to 4 rounds (ablation)language-only NSP+MLM · VisDial v1.0 val (language-only)NDCG: +4.56 on NDCG, +8.54 on MRRImproved
increase dialog history from 4 to 10 rounds (ablation)language-only NSP+MLM · VisDial v1.0 val (language-only)NDCG: +2.16 on NDCG, +1.57 on MRRImproved
add visual stream initialized from language-only weightsViLBERT · VisDial v1.0 valNDCG: 57.26 -> 62.64Improved
pretrain on CC before VisDial finetuning (vs L-only init) (ablation)ViLBERT · VisDial v1.0 valNDCG: 62.64 -> 60.80Hurt
add VQA pretraining after CC (ablation)ViLBERT · VisDial v1.0 valNDCG: 60.80 -> 64.94Improved
finetune on dense annotations with CE lossViLBERT w/ CC+VQA · VisDial v1.0 valNDCG: 64.94 -> 75.24 NDCG; 69.10 -> 52.22 MRRMixed
add NSP loss to dense CE finetuning (ablation)ViLBERT dense CE · VisDial v1.0 valMRR: 52.22 -> 65.88 MRR; 75.24 -> 69.24 NDCGImproved
zero-shot BERT (Wikipedia+BooksCorpus) with no VisDial finetuning (ablation)language-only NSP+MLM · VisDial v1.0 valNDCG: 57.26 -> 11.63Hurt
finetune only loss heads with other layers frozen (ablation)language-only NSP+MLM · VisDial v1.0 valNDCG: 11.63 -> 19.69Improved
add inconsistency loss in same batches (ablation)language-only NSP+MLM · VisDial v1.0 valNDCG: 57.26 -> 56.22Hurt
use jumbled inconsistency loss with NSP (ablation)language-only NSP · VisDial v1.0 valNDCG: 55.66 -> 55.83No clear effect

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