Towards Reliable LLM Systems with Neuro-Symbolic Assistance | Yu Feng
💬 Discord: https://discord.gg/h8NVzwnysW 💻 GitHub: https://github.com/centaurinstitute 🤝 LinkedIn: https://www.linkedin.com/company/centaur-ai-institute/ 📢 New Initiative: Neuro-Symbolic Agentic Protocol ⭐ Star us on GitHub https://github.com/centaurinstitute/neuro-symbolic-agentic-protocol — 📘 This week, Yu Feng will present two recent works that show how neuro-symbolic methods can complement LLMs by combining their flexibility with the rigor of probabilistic inference and formal verification to make LLM-based systems more trustworthy. Trustworthiness and reliability have become central challenges as LLMs have become increasingly deployed in real-world applications. 📄 First, she will introduce BIRD, a framework that makes use of LLM-generated factors to construct a Bayesian network that is then used to improve probabilistic decision-making. BIRD uses LLMs to identify factors, structure the underlying graph, and guide the approximation of conditional probability tables. The resulting Bayesian network performs correct probabilistic reasoning, leading to a more robust and trustworthy reasoning pipeline. Then she will present VeriCoT, a neuro-symbolic automated verification method for evaluating the logical validity and groundedness of chain-of-thought reasoning. VeriCoT extracts structured logical arguments from natural language reasoning and formalizes them in first-order logic, enabling SMT solvers to verify logical consistency. It also identifies the premises of each reasoning step, supporting interpretable diagnosis of flawed reasoning as well as further fine-tuning to improve reasoning quality. 🎓 Yu Feng is a PhD student at the University of Pennsylvania advised by Prof. Dan Roth. Her research focuses on improving the reliability of LLM systems. Her work has appeared at venues including ICLR, ACL, EMNLP, and ECCV, and she co-organized the COLM 2025 workshop AI Agents: Capabilities and Safety. — 🗨️ Join us for an interactive session exploring Neuro-Symbolic AI, the emerging paradigm that blends the strengths of neural networks with symbolic reasoning. We will discuss how hybrid approaches can enhance generalization, interpretability, and reasoning, and how these methods are shaping the future of intelligent systems. Whether you’re a researcher, engineer, or simply curious about the cutting edge of AI, you’ll find an engaging space to learn, connect, and exchange ideas. Event Format: 🔍 Reading Group 🎙️ Panel Discussion 🧩 Tutorials/Workshop Series 💬 Research Roundtable 🛠️ Open-Source Project Review 📄 Paper Pitch 🤝 Collaboration Hour 🍵 Networking Mixer #AI #NeuroSymbolic #FutureOfAI