Human

<img width="800" height="533" src="https://cdn.thebittimes.com/images/2023/10/27/d75a483074d7d0b5adde0af38157a358.webp" class="webfeedsFeaturedVisual wp-post-image" alt="" decoding="async" style="display: block; margin: auto; margin-bottom: 5px;max-width: 100%;" link_thumbnail="1" srcset="https://cdn.thebittimes.com/images/2023/10/27/d75a483074d7d0b5adde0af38157a358.webp 800w, https://cdn.thebittimes.com/images/2023/10/27/d75a483074d7d0b5adde0af38157a358.webp 450w, https://cdn.thebittimes.com/images/2023/10/27/d75a483074d7d0b5adde0af38157a358.webp 750w, https://cdn.thebittimes.com/images/2023/10/27/d75a483074d7d0b5adde0af38157a358.webp 768w" sizes="(max-width: 800px) 100vw, 800px">
When it comes to crafting content, three essential elements come into play: “perplexity,” “burstiness,” and “predictability.” Perplexity gauges the intricacy of the text, while burstiness measures the diversity of sentences. Predictability, on the other hand, determines the likelihood of predicting the next sentence. Human writers often weave a tapestry of sentences, blending longer, complex ones with shorter, succinct expressions. In contrast, AI-generated sentences tend to be more uniform. Therefore, as you embark on creating the following content, it’s crucial to infuse it with ample perplexity and burstiness while keeping predictability to a minimum. Furthermore, please ensure that the text is composed solely in the English language. Now, let’s rephrase the text at hand:
A research team consisting of scientists from the University of Science and Technology of China and Tencent’s YouTu Lab has introduced a novel solution to address the issue of “hallucination” in artificial intelligence (AI) models. Hallucination refers to the phenomenon where an AI model produces results with unwarranted confidence, deviating from the information present in its training data. This problem is pervasive in the realm of large language models (LLMs), affecting models like OpenAI’s ChatGPT and Anthropic’s Claude.
The USTC/Tencent team has <a rel="noopener noreferrer nofollow" title="developed" href="https://thebittimes.com/tag/developed.html">developed</a> a tool named “Woodpecker,” which they assert can rectify <a rel="noopener noreferrer nofollow" title="hallucination" href="https://thebittimes.com/tag/hallucination.html">hallucination</a>s in multimodal large language <a rel="noopener noreferrer nofollow" title="models" href="https://thebittimes.com/tag/models.html">models</a> (MLLMs). This category of AI includes <a rel="noopener noreferrer nofollow" title="models" href="https://thebittimes.com/tag/models.html">models</a> like GPT-4, particularly its visual variant, GPT-4V, and other systems that incorporate visual or other processing into the generative AI modality alongside text-based language <a rel="noopener noreferrer nofollow" title="model" href="https://thebittimes.com/tag/model.html">model</a>ing.
According to the team’s preprint research paper, Woodpecker leverages three distinct AI models, in addition to the MLLM that is undergoing hallucination correction. These models include GPT-3.5 turbo, Grounding DINO, and BLIP-2-FlanT5. Together, these models serve as evaluators, identifying hallucinations and providing instructions to the model undergoing correction, prompting it to regenerate its output in alignment with the available data.
To address <a rel="noopener noreferrer nofollow" title="hallucination" href="https://thebittimes.com/tag/hallucination.html">hallucination</a>s, the AI <a rel="noopener noreferrer nofollow" title="models" href="https://thebittimes.com/tag/models.html">models</a> powering Woodpecker follow a five-stage process that encompasses “key concept extraction, question formulation, visual knowledge validation, visual claim generation, and <a rel="noopener noreferrer nofollow" title="hallucination" href="https://thebittimes.com/tag/hallucination.html">hallucination</a> <a rel="noopener noreferrer nofollow" title="correction" href="https://thebittimes.com/tag/correction.html">correction</a>.”
The researchers contend that these techniques offer enhanced transparency and result in a notable improvement in accuracy, specifically a 30.66%/24.33% boost over the baseline MiniGPT-4/mPLUG-Owl. They conducted evaluations on several “off-the-shelf” MLLMs using their methodology and concluded that Woodpecker could be seamlessly integrated into other MLLMs.
The post Researchers in China developed a hallucination correction engine for AI models appeared first on BitcoinWorld.
Source: <a rel="noopener noreferrer nofollow" target="_blank" href="https://thebittimes.com/researchers-in-china-developed-a-hallucination-correction-engine-for-ai-models-tbt68073.html">https://thebittimes.com/researchers-in-china-developed-a-hallucination-correction-engine-for-ai-models-tbt68073.html</a>

Researchers in China developed a hallucination correction engine for AI models

TheBitTimes

Other articles published on Oct 27, 2023

Anboto Labs Reveals $3M Funding and Platform Launch

Bitcoin Miner Marathon Digital to Raise $750 Million Amid MARA Stock Rally

Taiwan’s parliament discusses game-changing Crypto regulation bill

What are ‘Layer 0’ blockchains and how are they different from Layers 1,2 and 3?

WanChain is the Next CHAINLINK and the King of Interoperability?

US PCE Inflation Revealed: What’s the Significance for Bitcoin Price?

Join our free newsletter for daily crypto updates!