AI Paper Atlas

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

Q&A Prompts: Discovering Rich Visual Clues through Mining Question-Answer Prompts for VQA requiring Diverse World Knowledge

Haibo Wang, Weifeng Ge

ECCV 2024 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 9 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 Q&A Prompts to InstructBLIP-7BInstructBLIP (7B) · A-OKVQA valaccuracy (%): 64.0 -> 69.4Improved
add Q&A Prompts to InstructBLIP-7BInstructBLIP (7B) · OK-VQAaccuracy (%): 62.1 -> 64.3Improved
add Q&A Prompts to different MLLMsMiniGPT-4, LLaVA-1.5, InstructBLIP-T5 · A-OKVQAaccuracy (%): +5.5 to +7.0Improved
train the VQG model instead of prompting an untrained MLLM for questions (ablation)Q&A Prompts · A-OKVQA valaccuracy (%): 61.9 -> 69.4Improved
untrained-VQG Q&A prompts versus no Q&A at all (ablation)InstructBLIP (7B) · A-OKVQA valaccuracy (%): 64.0 -> 61.9Hurt
train VQG on a mismatched dataset (OK-VQA instead of A-OKVQA) for A-OKVQA (ablation)Q&A Prompts · A-OKVQA valaccuracy (%): 69.4 -> 67.1Hurt
encode Q&A with visual-aware prompting module instead of prepending as text (ablation)Q&A Prompts · A-OKVQA valaccuracy (%): 66.6 -> 69.4Improved
use Q&A pairs instead of captions or tags as prompts (ablation)Q&A Prompts (prepend) · A-OKVQA valaccuracy (%): 64.2 -> 66.6Improved
increase number of Q&A pairs from 8 to 16 (ablation)Q&A Prompts · A-OKVQA valaccuracy (%): 69.4 -> 69.2No clear effect
increase number of Q&A pairs from 1 to 8 (ablation)Q&A Prompts · A-OKVQA valaccuracy (%): 66.2 -> 69.4Improved
add visual gated fusion to the prompting module (ablation)visual-aware prompting module · A-OKVQA valaccuracy (%): 68.6 -> 69.4Improved
add resampler decoder to compress prompts to 32 tokens (ablation)visual-aware prompting module · A-OKVQA valaccuracy (%): 68.1 -> 69.4Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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