Reproduce “ClaimVer: Explainable Claim-Level Verification and Evidence Attribution of Text Through Knowledge Graphs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Preetam Prabhu Srikar Dammu, Himanshu Naidu, Mouly Dewan, YoungMin Kim, Tanya Roosta, Aman Chadha, Chirag Shah
ClaimVer introduces a framework for explainable, claim-level verification and evidence attribution of text by integrating Knowledge Graphs like Wikidata with fine-tuned large language models. The system accurately decomposes input text into individual claims, provides specific factual evidence, and generates human-centric rationales, enabling precise identification of attributable, extrapolatory, or contradictory information.
Key insights
- Leveraging Knowledge Graphs provides a structured, consolidated source of truth for evidence, overcoming limitations of distributed, one-to-one reference-based fact-checking.
- Integrating claim decomposition and attribution into a single LLM query significantly enhances efficiency, avoiding the exponential computational costs of separate processing steps.
- The generated human-centric rationales and specific evidence (KG triplets) directly address the need for explainable AI, fostering user trust in automated fact-checking systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.