FEVER

Advance automated fact-checking and verifiable knowledge extraction using research-grade datasets and shared tasks for NLP researchers and AI developers.

Free fact checking

No paid plan

Free

About FEVER

FEVER (Fact Extraction and VERification) is a specialized research initiative and workshop series dedicated to advancing the state of automated fact-checking (AFC). As information on the web grows exponentially, the gap between unstructured text and verifiable knowledge has widened, leading to the proliferation of misinformation. FEVER addresses this by providing a standardized framework and venue for researchers to develop systems that can accurately extract facts and verify claims against reliable evidence. By providing high-quality benchmark datasets, the platform helps standardize the evaluation of models designed to distinguish between factual information and unreliable claims. The platform operates through annual workshops and Shared Tasks where participants compete to build models that solve specific verification problems. For the upcoming Ninth FEVER Workshop (FEVER9), the focus has shifted toward multimodal verification through the AVerImaTeC task. This involves verifying image-text claims using evidence gathered from the web, requiring models to integrate visual and textual understanding. Beyond the annual tasks, the project maintains a collection of datasets that serve as benchmarks for recognizing textual entailment, question answering, and argumentation mining. These resources are essential for training the next generation of Large Language Models to be more grounded and factually accurate. This tool is primarily intended for academic researchers, data scientists, and computational linguists specializing in Natural Language Processing (NLP). It provides the necessary infrastructure—specifically high-quality labeled datasets and a peer-reviewed community—to test new algorithms for information retrieval and claim validation. It is also highly relevant for organizations developing trust and safety tools or news verification software looking to implement cutting-edge AFC techniques. Because the datasets are derived from real-world web content, they offer a rigorous testing ground for models intended for production use in media and technology sectors. Unlike generic AI platforms, FEVER is a community-driven academic effort co-located with major conferences like EACL. Its focus on verifiable knowledge sets it apart; rather than just generating text, the emphasis is on the provenance of information. By bridging the gap between free-form web content and structured knowledge sources, FEVER pushes the boundaries of how AI can combat misinformation at scale. Through its ongoing workshops, it continues to define the benchmarks used by the global AI community to ensure information integrity.

FEVER pros & cons

Pros

Provides high-quality, research-validated datasets for fact-checking model training.
Focuses on the difficult problem of multimodal image-text claim verification.
Co-located with top-tier NLP conferences like EACL, ensuring high academic standards.
Active community support through a dedicated Slack channel and annual shared tasks.

Cons

Primarily designed for researchers, which may be difficult for non-technical users to navigate.
Participation and submission windows are bound by strict academic conference timelines.
Requires high-level expertise in NLP and machine learning to utilize datasets effectively.

FEVER use cases

NLP researchers can use FEVER datasets to benchmark their latest models for recognizing textual entailment and fact verification.
Trust and safety teams can adapt the shared task methodologies to improve automated detection of multimodal misinformation.
Data scientists developing search engines can implement the verifiable knowledge extraction techniques to provide evidence-backed answers.
Academic students can participate in shared tasks to gain hands-on experience with real-world fact-checking and multimodal data challenges.

FEVER features

information retrieval integration
argumentation mining tasks
textual entailment evaluation
claim-evidence alignment
peer-reviewed workshop venue
averimatec dataset
multimodal verification tasks
automated fact-checking benchmarks

FEVER pricing

Is FEVER free? Yes, FEVER has a free plan.

Research Access

Free

Access to FEVER datasets
Participation in Shared Tasks
Workshop paper submission
AVerImaTeC dataset access
Slack community access
Standardized evaluation metrics

FEVER FAQs

What is the primary focus of the FEVER9 shared task?

The Ninth FEVER Workshop introduces the AVerImaTeC task, which focuses on automated verification of claims involving both images and text. Participants must develop systems that can verify these multimodal claims using evidence retrieved from the live web.

Who can participate in the FEVER workshops?

Researchers, students, and practitioners in the fields of NLP, ML, and information retrieval are encouraged to participate. Submissions can range from workshop papers to specific results from the shared tasks.

Are datasets available for public use?

Yes, the platform provides specialized datasets like AVerImaTeC for training and evaluating fact-checking models. These datasets are designed to move beyond structured sources to handle complex, free-form web information.

How are submissions reviewed for the workshop?

FEVER follows a formal academic review process with specific deadlines for paper submissions and camera-ready versions. Accepted papers are presented at the workshop, which is co-located with the EACL conference.

Ratings & reviews

No reviews yet. Be the first to share how FEVER worked for you.