AI Paper Atlas

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

Scallop: From Probabilistic Deductive Databases to Scalable Differentiable Reasoning

Jiani Huang, Ziyang Li, Binghong Chen, Karan Samel, Mayur Naik, Le Song, Xujie Si

NeurIPS 2021 · Vision-language and multimodal › Visual question answering · approach: Vision-language model VQA evaluation · 17 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
Scallop top-k differentiable reasoning versus Neural Module NetworksVQAR · VQAR 5000 CallRecall@5: 71.80% -> 84.22%Improved
Scallop versus LXMERTVQAR · VQAR 5000 CallRecall@5: 62.56% -> 84.22%Improved
remove differentiability, use REINFORCE with discrete scene graph sampling (Datalog-RL) (ablation)Scallop · VQAR generalization to harder questionsRecall@5Hurt
increase k from 1 to 5 in training (ablation)Scallop · sum3convergence stabilityImproved
increase k from 1 to 15 at inference (ablation)Scallop · sum3inference time (s): 0.05 -> 6.15Hurt
use smaller ktest than ktrain=10 at test time (ablation)Scallop · sum3accuracy difference vs ktest=10No clear effect
Scallop top-k versus exact DeepProbLogsynthetic MNIST tasks · 6 synthetic tasksaccuracyNo 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