PhD Candidate @ AI Institute | Artificial Intelligence | Neuroscience | Neurosymbolic AI | Knowledge Graph | Knowledge Infused Learning | Natural Language Processing
Our paper titled "Neurosymbolic AI Approach to Attribution in Large Language Models" 𝗵𝗮𝘀 𝗯𝗲𝗲𝗻 𝗮𝗰𝗰𝗲𝗽𝘁𝗲𝗱 for publication in 𝗜𝗘𝗘𝗘 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁 𝗦𝘆𝘀𝘁𝗲𝗺𝘀! 🎉 This work addresses a pressing challenge in large language models (LLMs): 𝗮𝘁𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗲𝗻𝘀𝘂𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗳𝗮𝗰𝘁𝘂𝗮𝗹 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗼𝗳 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱 𝗰𝗼𝗻𝘁𝗲𝗻𝘁. Current citation tools (e.g., Perplexity.ai, Bing-integrated LLMs) attempt to ground responses with real-time searches and citations. Still, they often face challenges like hallucinations, biases, and the use of unreliable sources. Our proposed solution? Neurosymbolic AI (NesyAI)! 🤖✨ By integrating LLMs with structured, symbolic reasoning, NesyAI enables more 𝘁𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝘁, 𝗶𝗻𝘁𝗲𝗿𝗽𝗿𝗲𝘁𝗮𝗯𝗹𝗲, 𝗮𝗻𝗱 𝗿𝗲𝗹𝗶𝗮𝗯𝗹𝗲 𝗮𝘁𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 for LLM outputs. This approach provides dynamic and accurate reasoning while mitigating biased or incomplete sources. With NesyAI, we can enhance the reliability and adaptability of LLMs, moving toward AI systems that offer more trustworthy, evidence-based responses. 🌟 🔗 Read the abstract and link for the full paper (https://lnkd.in/eVHSVPJX) Revathy Venkataramanan Amit Sheth Artificial Intelligence Institute of South Carolina #ArtificialIntelligence #NeurosymbolicAI #LLM #Attribution #Research #IEEE #MachineLearning #AIethics #TrustworthyAI
