Researchers in China developed a hallucination correction engine for AI models

A team of scientists from the University of Science and Technology of China and Tencent’s YouTu Lab have developed a tool to combat “hallucination” by artificial intelligence (AI) models. 

Hallucination is the tendency for an AI model to generate outputs with a high level of confidence that don’t appear based on information present in its training data. This problem permeates large language model (LLM) research, and its effects can be seen in models such as OpenAI’s ChatGPT and Anthropic’s Claude.

How To Get Free Crypto  

The USTC/Tencent team developed a tool called “Woodpecker” that they claim is capable of correcting hallucinations in multimodal large language models (MLLMs). 

This subset of AI involves models such as GPT-4 (especially its visual variant, GPT-4V) and other systems that roll vision and/or other processing into the generative AI modality alongside text-based language modeling. 

According to the team’s preprint research paper, Woodpecker uses three separate AI models, apart from the MLLM being corrected for hallucinations, to perform hallucination correction. 

These include GPT-3.5 turbo, Grounding DINO and BLIP-2-FlanT5. Together, these models work as evaluators to identify hallucinations and instruct the model being corrected to regenerate its output in accordance with its data. 

Coinbase Banner  

In each of the above examples, an LLM hallucinates an incorrect answer (green background) to prompting (blue background). The corrected Woodpecker responses are shown with a red background. Source: Yin, et. al., 2023

To correct hallucinations, the AI models powering Woodpecker use a five-stage process that involves “key concept extraction, question formulation, visual knowledge validation, visual claim generation, and hallucination correction.”

Related: Humans and AI often prefer sycophantic chatbot answers to the truth — Study

The researchers claim these techniques provide additional transparency and “a 30.66%/24.33% improvement in accuracy over the baseline MiniGPT-4/mPLUG-Owl.” They evaluated numerous “off the shelf” MLLMs using their method and concluded that Woodpecker could be “easily integrated into other MLLMs.”

Go to Source
Author: Tristan Greene

Recommended Crypto Services, Products and Strategies:

The first thing any crypto investor needs is is a reliable and secure Crypto Wallet.  Whether you’re looking for an online wallet, hardware wallet, desktop or mobile wallet, Crypto Renegade provides you with all the Best Crypto Wallets in each category.

FreeBitcoin Banner  

Best Crypto Wallets Banner

When you’re ready to buy more crypto, or exchange your coins for others, Crypto Renegade’s list of the Best Crypto Exchanges has you covered.  The Crypto Exchanges recommended here offer everything from simplicity and convenience to advanced trading platforms and profit sharing. 

Best Crypto Exchanges Banner

If you want to learn more about the methods and tools that can be used to find Great Crypto Projects, then be sure to check out Crypto Renegade’s strategy for How To Find The Best Cryptocurrency.

Crypto Strategy Banner

For those people that don’t have any money to invest right now, or just want to understand the technology a bit more, you’ll definitely want to check out Crypto Renegade’s Free Crypto Strategy and start collecting Free Coins today!

Free Crypto Banner

What do you think about cryptocurrency? Do you have any questions about it? Be sure to leave a comment below.

You May Also Like

More From Author

This site uses Akismet to reduce spam. Learn how your comment data is processed.