VITA-MLLM/Woodpecker
✨✨Woodpecker: Hallucination Correction for Multimodal Large Language Models
A training-free post-processing pipeline that corrects hallucinations in MLLM outputs through five stages: concept extraction, question formulation, visual validation with GroundingDINO, claim generation, and correction. Integrates with LLaVA, mPLUG-Owl, Otter, and MiniGPT-4, leveraging spaCy for NLP and GPT-4V for evaluation, achieving 30.66% accuracy improvements on POPE benchmarks without model retraining.
650 stars. No commits in the last 6 months.
Stars
650
Forks
30
Language
Python
License
—
Category
Last pushed
Dec 23, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/VITA-MLLM/Woodpecker"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
vectara/hallucination-leaderboard
Leaderboard Comparing LLM Performance at Producing Hallucinations when Summarizing Short Documents
PKU-YuanGroup/Hallucination-Attack
Attack to induce LLMs within hallucinations
NishilBalar/Awesome-LVLM-Hallucination
up-to-date curated list of state-of-the-art Large vision language models hallucinations...
Amirhosein-gh98/Gnosis
Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits
MemTensor/HaluMem
HaluMem is the first operation level hallucination evaluation benchmark tailored to agent memory systems.