MILVLG/prophet

Implementation of CVPR 2023 paper "Prompting Large Language Models with Answer Heuristics for Knowledge-based Visual Question Answering".

43
/ 100
Emerging

Employs a two-stage framework that first trains a vanilla VQA model (MCAN) on knowledge-based datasets to extract answer candidates and answer-aware examples as heuristics, then uses these heuristics to prompt GPT-3 for improved reasoning. Supports multi-GPU training with PyTorch on OK-VQA and A-OKVQA datasets, leveraging pre-computed image features and extensive pre-training on VQA v2 before task-specific fine-tuning. Achieves SOTA performance by combining supervised VQA models with large language models through strategic prompt engineering.

279 stars. No commits in the last 6 months.

Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 15 / 25

How are scores calculated?

Stars

279

Forks

28

Language

Python

License

Apache-2.0

Last pushed

Jun 14, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/prompt-engineering/MILVLG/prophet"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.