Repository for the SymTrust: Symbolic Methods for Trustworthy and Verifiable Reasoning ACL workshop.
Workshop website: https://hlr.github.io/symtrust/
Large language models are now used in many important areas, making it essential to know when their outputs can be trusted. However, correct answers may still follow flawed reasoning. Verifying how models reach their answers has therefore become necessary. SymTrust asks a central question: how can we determine whether an answer is supported by reliable reasoning rather than merely appearing correct?
One promising direction is to use methods that make reasoning easier to inspect and verify. A growing body of work explores symbolic and structured methods, including rules, constraints, programs, graphs, knowledge bases, and diagrams. These methods can improve model reasoning and express relevant information and relations in forms that support evaluation and verification.
SymTrust focuses on the connection between trustworthy AI and symbolic and structured methods. The workshop asks how these methods can help improve and verify model reasoning, and build more reliable, understandable, and trustworthy AI systems. We welcome researchers working across domains, modalities, and reasoning tasks to discuss new methods, evaluations, benchmarks, applications, and open challenges at the intersection of trustworthy AI, symbolic reasoning, and structured representations.