PaperlayerView exact revision on arXiv

Source-linked AI summary

Overview of ROMCIR 2026: The 6th Workshop on Reducing Online Misinformation through Credible Information Retrieval

Marcos Fernández-Pichel, Marinella Petrocchi, Kevin Roitero, Marco Viviani

arXiv:2609.05684v1cs.IR

TL;DR

Online information disorder makes it important for information retrieval to provide relevant, truthful, and factually accurate information. This overview presents ROMCIR’s aims and accepted submissions across three research areas; 5 of 10 submissions were accepted, with work spanning fact-checking, generative-model reliability, and human-centered evaluation.

  • Problem

    Online information disorder includes varied forms of communication pollution, while retrieval systems face challenges involving truthfulness, explainability, and reliable evaluation.

    Source: §1 · pp. 1–2

  • Method

    The overview organizes ROMCIR 2026’s accepted work around explainable multimodal fact-checking, generative-model reliability, and datasets and human-centered evaluation.

    Source: §4 · pp. 3–4

  • Results

    5 of 10 submissions were accepted, a 50% acceptance rate, spanning explainable multimodal fact-checking, generative-model reliability, and human-centered evaluation.

    Source: §4 · p. 3

  • Takeaways & Limitations

    ROMCIR aims to engage the information retrieval community in solutions beyond conventional misinformation detection, including truthful and explainable search and responsible generative-model integration.

    Source: §1 · p. 2

Abstract

from arXiv · show

In the digital online ecosystem, we are surrounded by distinct forms of information pollution, posing significant threats to both individuals and society. Fake news, for instance, wields power to sway public opinion on matters of politics and finance. Deceptive reviews can either bolster or tarnish the reputation of businesses, while unverified medical advice may steer people toward harmful health practices. In light of this challenging landscape, it has become imperative to ensure that users have access to both topically relevant and factually accurate information that does not warp their perception of reality, and there has been a surge of interest in various strategies to combat misinformation through different contexts and multiple tasks. The purpose of the ROMCIR Workshop, for some years now, is precisely that of engaging the Information Retrieval community to explore potential solutions that extend beyond conventional misinformation detection approaches. Key objectives include identifying subjective and objective factors associated with information credibility and truthfulness, respectively, and integrating such factors as fundamental dimensions of relevance within IR Systems (IRSs), achieving early detection of misinformation, and ensuring that the search results retrieved are not only truthful but also explainable to the users of IRSs. Moreover, it is essential to evaluate the role of generative models such as Large Language Models (LLMs) in inadvertently amplifying misinformation problems, and how they can be used to support IRSs, together with the contribution that the human-in-the-loop paradigm can have in this context.

1. Introduction

ROMCIR 2026 focuses on information retrieval that serves users relevant information while also supporting truthfulness and factual accuracy. The introduction frames this as a complex problem shaped by varied content, platforms, intentions, and digital environments.

1. Introduction

  • ROMCIR 2026 addresses access to topically relevant and truthful, reliable, or factually accurate information amid online information disorder.The introduction uses “information disorder” for communication pollution ranging from misinformation shared through ignorance to intentional falsehood.Source: §1 · p. 1
  • The challenge spans heterogeneous content, dissemination platforms, user intentions, and digital features such as filter bubbles and echo chambers.These structural features can reinforce pre-existing beliefs and limit exposure to diverse perspectives.Source: §1 · pp. 1–2
  • Emerging concerns include explainability, truthfulness in user-generated content, crowdsourced verification, generative models, data confidentiality, and reliable IR evaluation.The introduction presents these as dimensions of designing effective and reliable experimental evaluation paradigms for information retrieval systems.Source: §1 · p. 2

2. Aim and Topics of Interest

ROMCIR 2026 spans misinformation and trustworthy information retrieval across content types, platforms, and user goals. Its topics also encompass AI-related challenges and methods for assessing or retrieving truthful information.

2. Aim and Topics of Interest

  • The workshop covers misinformation and trustworthy information retrieval across web pages, news, reviews, medical data, social media, microblogs, and Q&A systems.User goals include detecting falsehoods and retrieving truthful information.Source: §2 · p. 2
  • AI-related topics include explaining search results, assessing truthfulness in AI-generated content, and integrating generative models into information retrieval systems.These topics are included alongside more established misinformation and disinformation concerns.Source: §2 · p. 2
  • The topic list includes fact-checking, credibility assessment, crowdsourcing, human-in-the-loop detection, reliable retrieval, and misinformation’s societal and community dimensions.It also includes query reformulation, verification strategies, privacy, security, propaganda, and trust and reputation.Source: §2 · p. 2

3. Keynote Speech

The keynote examines misinformation detection’s progress and unresolved challenges, including the gap between technical advances and societal impact. It calls for greater engagement with practitioners and attention to outcomes beyond benchmarks.

3. Keynote Speech

  • Tommaso Caselli delivered the keynote on recent advances in misinformation detection, its challenges, and societal impact.His research includes natural language processing, event extraction, framing, and detecting and mitigating hate speech and misinformation.Source: §3 · pp. 2–3
  • The talk situates the rise of misinformation-detection research around increased public attention and two major events: the 2016 US Presidential Election and the COVID-19 pandemic.These events accelerated research into detection tools and benchmark datasets.Source: §3 · p. 2
  • The keynote identifies a gap between technical progress in misinformation detection and societal impact, alongside the unresolved question of helping people correct established false beliefs.It argues that researchers should engage more deeply with practitioners and measure outcomes rather than benchmarks.Source: §3 · p. 3

4. Submissions

ROMCIR 2026 received 10 submissions and accepted 5, spanning explainable multimodal fact-checking, generative-model reliability, and human-centered evaluation. The accepted work includes new resources and evaluation approaches as well as findings about model and crowd limitations.

4. Submissions

  • 5 of 10 submissions were accepted, a 50% acceptance rate, with work focused on explainable multimodal fact-checking, generative-model reliability, and human-centered evaluation.Accepted authors were affiliated with universities in six countries: France, Italy, Spain, Turkey, the United Kingdom, and Vietnam.Source: §4 · p. 3
  • Two accepted papers address multimodal fact-checking: one surveys explainable automated fact-checking, while another introduces a near-duplicate image-retrieval benchmark.The benchmark contains 2,473 images in near-duplicate clusters with professional claim reviews; the authors explore contrastive fine-tuning with simulated image modifications.Source: §4 · p. 3
  • The generative-model submissions examine RAG reliability under clean, poisoned, or mixed evidence and reassess contextual-hallucination benchmarks through human adjudication.Yigit-Sert identifies breaking points using Parametric Override Rate and Confidence Inflation; Atasoy et al. report that LLMs capture inconsistencies missed by single-pass human annotation.Source: §4 · pp. 3–4
  • Crowdsourcing provides a scalable screening signal for audiovisual deepfakes, but reliable identification of manipulation modality remains challenging.Soprano et al. conclude that crowd assessment is most effective as a preliminary step in broader, multi-step verification.Source: §4 · p. 4

5. Past Editions

ROMCIR’s first five editions, held annually from 2021 to 2025, brought discussion and innovative work on information disorder and Information Retrieval to ECIR; their proceedings are freely accessible.

  • ROMCIR’s first five editions ran annually from online (2021) through Stavanger (2022), Dublin (2023), Glasgow (2024), and Lucca (2025).Source: §5 · p. 4
  • The editions fostered discussion and presentation of innovative work on open issues in information disorder and Information Retrieval.Source: §5 · p. 4
  • Workshop papers appeared in freely accessible CEUR Workshop Proceedings, and the website provides updated edition information and materials including keynote slides and family photos.Source: §5 · p. 4

6. Workshop Organization

The workshop’s realization relied on four main organizers, a Proceedings and Publicity Chair, and a Program Committee, whose members’ biographies describe expertise across information retrieval, misinformation, and related areas.

  • Workshop organization relied on four main organizers, a Proceedings and Publicity Chair, and Program Committee members.Source: §6 · p. 4
  • Marcos Fernández-Pichel’s research focuses on online health-information credibility, Information Retrieval, and NLP applied to mental health.Source: §6 · p. 4
  • Marinella Petrocchi studies fake-news and fake-account detection and automated ranking of online news-media reputability; she leads the Humane project at CNR.Source: §6 · p. 5
  • Kevin Roitero’s research spans Artificial Intelligence, NLP, Information Retrieval, and crowdsourcing, alongside academic and industry collaborations.Source: §6 · p. 5
  • Marco Viviani’s research includes Social Computing, Information Retrieval, trust and reputation management, and user modeling.Source: §6 · p. 5
  • Riccardo Lunardi served as Proceedings and Publicity Chair.Source: Proceedings and Publicity Chair · p. 5
Loading 2609.05684v1…