@prefix this: . @prefix sub: . @prefix np: . @prefix dct: . @prefix pav: . @prefix rdf: . @prefix owl: . @prefix rdfg: . @prefix dce: . @prefix xsd: . @prefix rdfs: . @prefix prov: . @prefix npx: . sub:Head { this: a np:Nanopublication; np:hasAssertion sub:assertion; np:hasProvenance sub:provenance; np:hasPublicationInfo sub:pubinfo . } sub:assertion { a prov:Entity; dct:title "KG-CRAFT: Knowledge Graph-based Contrastive Reasoning with LLMs for Enhancing Automated Fact-checking"; ; , , , , , , , , , , , , , , , , , , . a ; rdfs:comment "CO-GAT (ELECTRA) is discussed as a competing method for claim verification that applies graph attention over scientific evidence, utilizing a pre-trained ELECTRA encoder."; rdfs:label "CO-GAT (ELECTRA)" . a ; rdfs:comment "CofCED is discussed as a traditional automated fact-checking (AFC) baseline that introduced a hierarchical encoder with cascaded evidence selectors to process reports from multiple sources for claim verification."; rdfs:label "CofCED" . a ; rdfs:comment "CorXFact is discussed as an LLM-based fact-checking approach that simulates human fact-checking principles by analyzing claim-evidence correlations."; rdfs:label "CorXFact" . a ; rdfs:comment "dEFEND is discussed as a traditional automated fact-checking (AFC) baseline method that employs a sentence-comment co-attention network to jointly analyze news content and user comments for claim verification."; rdfs:label "dEFEND" . a ; rdfs:comment "FactL-LaMA is discussed as a specialized LLM-based fact-checking method that combines instruction-following with external evidence retrieval, implemented as a Low-Rank Adaptation (LoRA) fine-tuned Llama 2 7B."; rdfs:label "FactL-LaMA" . a ; rdfs:comment "FactL-LaMA-Know is discussed as a specialized LLM-based fact-checking method that augments the FactL-LaMA model with external knowledge sources to improve fact-checking accuracy."; rdfs:label "FactL-LaMA-Know" . a ; rdfs:comment "FIRE is discussed as a recent LLM-based fact-checking approach that proposes an iterative verification method aiming to improve scalability and efficiency in assessing claims."; rdfs:label "FIRE" . a ; rdfs:comment "GenFE is discussed as a traditional automated fact-checking (AFC) baseline that performs joint optimization of veracity prediction and explanation generation, showing improved classification and output quality."; rdfs:label "GenFE" . a ; rdfs:comment "GenFE-MT is discussed as a traditional automated fact-checking (AFC) baseline that performs joint optimization of veracity prediction and explanation generation, showing improved classification and output quality."; rdfs:label "GenFE-MT" . a ; rdfs:comment "GraphCheck is discussed as a recent LLM-based verifier that incorporates lightweight graph signals and instruction-style prompting for fact-checking."; rdfs:label "GraphCheck" . a ; rdfs:comment "GraphFC is discussed as a competing method for claim verification that encodes the report structure with an explicit graph and graph reasoning components."; rdfs:label "GraphFC" . a ; rdfs:comment "IKA is discussed as an LLM-based fact-checking framework that builds a graph of positive and negative examples from labelled data to enhance fact-checking and explanation capabilities."; rdfs:label "IKA" . a ; dct:subject ; rdfs:comment "KG-CRAFT is a novel method that enhances LLM performance in automated fact-checking. It leverages knowledge graphs to construct structured contrastive questions from claims and reports, which then guide an LLM to generate evidence-based summaries for veracity assessment during the inference stage. This process enables LLMs to access and reason with contextualized knowledge more effectively."; rdfs:label "KG-CRAFT"; . a ; rdfs:comment "L-Defense is discussed as a specialized LLM-based fact-checking framework that leverages the wisdom of crowds by partitioning evidence into competing narratives for robust claim verification using LLMs like Llama 2 7B and ChatGPT 3.5 Turbo."; rdfs:label "L-Defense" . a ; rdfs:comment "MLA (RoBERTa) is discussed as a competing method for claim verification that uses a sequence inference model with self-attention at token and sentence levels, powered by a pre-trained RoBERTa encoder."; rdfs:label "MLA (RoBERTa)" . a ; rdfs:comment "MULTIVERS is discussed as a competing method for claim verification that leverages a multi-task learning approach for scientific verification and rationale extraction from abstracts."; rdfs:label "MULTIVERS" . a ; rdfs:comment "PACAR is discussed as a competing prompting-based approach for fact verification that aggregates evidence with consistency-oriented reasoning using large language models."; rdfs:label "PACAR" . a ; rdfs:comment "ProgramFC is discussed as a competing method for claim verification that formulates fact verification as a programmatic pipeline with modular steps for evidence usage and decision-making."; rdfs:label "ProgramFC" . a ; rdfs:comment "SBERT-FC is discussed as a traditional automated fact-checking (AFC) baseline method that uses SBERT to encode claims and documents, performing verification via supervised classifiers and providing explainability analysis."; rdfs:label "SBERT-FC" . a ; rdfs:comment "TELLER is discussed as an LLM-based fact-checking framework that is a dual-system approach integrating human expertise with LLM reasoning to emphasize trustworthiness."; rdfs:label "TELLER" . } sub:provenance { sub:assertion prov:wasAttributedTo ; prov:wasDerivedFrom . } sub:pubinfo { this: dct:created "2026-03-13T16:08:16.428Z"^^xsd:dateTime; dct:creator ; npx:hasNanopubType ; npx:supersedes ; rdfs:label "LLM-KG assessment for paper 10.48550/arXiv.2601.19447" . sub:sig npx:hasAlgorithm "RSA"; npx:hasPublicKey "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAwNz2QK3SEifno78S7+48zUB0xpTex3mAzW73ZimHqNcdEMU5/apslrGrTHGFAt/Chocgo++r6JQp5ygY7NyJHGWdaIqnt85pjX4PbNfLAvapyUO00qZP34fY61w4eZ9UMtleWEsmZKRtQPyJ8ODl46i/rfPuZlcJGpM9Nmy5mpGWuepqIEvF4a/t7pLVeCEDFSYXT+yaiygt6ynIK5f7TtEDhZpeUf/Q74WhMPJXm4yTU/hqOX4IW+50kWHNArGGZwUaXwzyG6M3Zd6UMModryGkLqS4H/MSE3ZA1Ylnms7BfWLEXhMWlaKi6HRV4nGRDLhxVSi9LSRi3LWKLhNIIQIDAQAB"; npx:hasSignature "Vd3waP8e4zJi2Pk1git+rh/78E7EK+nKm7KEAnnDs8XqfEGUbt2GyK54kfWA7dh0twJrike2REjpnCoHnLjoZsxyJMw1x+Nh2uUzvkMFHKz+d3Ul23abkTUlEeKEZZa/BMByBoRg33e7eBuNpdER+3eVNLV4GUFoVF1DT1WaHIWTCesshHTez4lyb1gewCYhOycAJMlFLoMbG5HXXmQy5bfuY7qf7rDn7JKK0/qEnR0e0UUpe9Azq0aBx9LtZe6WfhnSHtmdlnvYeqYUqUfxPhor4cMeqoMgW5/UUse1XCauu7Na/C5ms35GBjZRfrlGrSUt5bp2zZIrCi0wXYUg0A=="; npx:hasSignatureTarget this:; npx:signedBy . }