Less is More: Improved Fact-Checking on AVeriTeC with Iterative Question Generation

• Lucas M. Fraga Fraga PUCRS
• Marcelo M. Delucis PUCRS
• Christian Mattjie PUCRS
• Pedro Wagner PUCRS
• Arthur Ibarra PUCRS
• Yasmin Cardozo Aguirre PUCRS
• Rodrigo C. Barros PUCRS / Kunumi Institute
• Lucas S. Kupssinskü PUCRS

DOI: https://doi.org/10.5753/eniac.2026.29752

Resumo

Automated open-web fact-checking decomposes a claim into questions, retrieves evidence, and produces a verdict. AVERITEC systems report gains from iterative question generation, but iteration is bundled with reranking, multi-agent coordination, or fine-tuned generators, leaving its contribution unquantified. We isolate iteration as the only variable in an otherwise fixed AVERITEC pipeline and compare it to single-pass generation across six instruction-tuned LLMs (4-8B). On the dev set, iterative generation improves accuracy in every case (1.8 to 21.2 pp) while more than halving queries per claim (4.96 → 2.38) and lowering abstention (0.424→ 0.221). Iteration alone accounts for the accuracy and decisiveness gains reported in prior AVERITEC pipelines.

Referências

Asai, A. et al. (2023). Self-rag: Learning to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations.

Bai, G. et al. (2024). Beyond efficiency: A systematic survey of resource-efficient large language models. arXiv preprint arXiv:2401.00625.

Barcelos, P., Parraga, O., Mussi, M. M., Fraga, L. M., Kupssinskü, L. S., and Barros, R. C. (2026). Latent fact-checking: Detecting misinformation through activation engineering.

Chen, J. et al. (2024). Complex claim verification with evidence retrieved in the wild. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 3569–3587.

Delucis, M. et al. (2025). Automated fact-checking in brazilian portuguese: Resources and baselines. In Anais do XVI Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana, pages 137–148. SBC.

Dharmavaram, A. et al. (2025). SANCTUARY: An efficient evidence-based automated fact checking system. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 247–257, Vienna, Austria. Association for Computational Linguistics.

Gao, L. et al. (2023). Precise zero-shot dense retrieval without relevance labels. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1762–1777.

Gemma Team, Mesnard, T., et al. (2024). Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295.

Gong, S. et al. (2026). Multi-sourced, multi-agent evidence retrieval for fact-checking. arXiv preprint arXiv:2603.00267.

Grattafiori, A. et al. (2024). The llama 3 herd of models. arXiv preprint arXiv:2407.21783.

Guo, D. et al. (2025a). Deepseek-r1 incentivizes reasoning in llms through reinforcement learning. Nature, 645(8081):633–638.

Guo, J. et al. (2025b). Stylebench: Evaluating thinking styles in large language models. arXiv preprint arXiv:2509.20868.

Hagström, L. et al. (2025). Language model re-rankers are fooled by lexical similarities. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 18–33, Vienna, Austria. Association for Computational Linguistics.

Hasanain, M. et al. (2022). Studying effectiveness of web search for fact checking. J. Assoc. Inf. Sci. Technol., 73(5):738–751.

Hong, S. et al. (2025). EMULATE: A multi-agent framework for determining the veracity of atomic claims by emulating human actions. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 179–183, Vienna, Austria. Association for Computational Linguistics.

Jiang, Z. et al. (2023). Active retrieval augmented generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 7969–7992, Singapore. Association for Computational Linguistics.

Jin, J. et al. (2025). Flashrag: A modular toolkit for efficient retrieval-augmented generation research. In Companion Proceedings of the ACM on Web Conference 2025, page 737–740, New York, NY, USA. Association for Computing Machinery.

Khaliq, M. A. et al. (2024). RAGAR, your falsehood radar: RAG-augmented reasoning for political fact-checking using multimodal large language models. In Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER), pages 280–296, Miami, Florida, USA. Association for Computational Linguistics.

Laitenberger, A. et al. (2025). Stronger baselines for retrieval-augmented generation with long-context language models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 32559–32569, Suzhou, China. Association for Computational Linguistics.

Liu, A. H. et al. (2026). Ministral 3. arXiv preprint arXiv:2601.08584.

Liu, J. et al. (2024a). FZI-WIM at AVeriTeC shared task: Real-world fact-checking with question answering. In Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER), pages 77–85, Miami, Florida, USA. Association for Computational Linguistics.

Liu, J. et al. (2024b). GProofT: A multi-dimension multi-round fact checking framework based on claim fact extraction. In Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER), pages 118–129, Miami, Florida, USA. Association for Computational Linguistics.

Malon, C. (2024). Multi-hop evidence pursuit meets the web: Team papelo at FEVER 2024. In Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER), pages 27–36. Association for Computational Linguistics.

Nakov, P. et al. (2021). Automated fact-checking for assisting human fact-checkers. arXiv preprint arXiv:2103.07769.

Nissen, L. et al. (2025). Medicine on the edge: Comparative performance analysis of on-device llms for clinical reasoning.

Putta, A. R. et al. (2025). ClaimCheck: Automatic fact-checking of textual claims using web evidence. In Proceedings of the 4th International Workshop on Knowledge-Augmented Methods for Natural Language Processing, pages 303–316, Albuquerque, New Mexico, USA. Association for Computational Linguistics.

Rashid, F. B. et al. (2025). Fathom: A fast and modular RAG pipeline for fact-checking. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 258–265, Vienna, Austria. Association for Computational Linguistics.

Rau, D. et al. (2024). BERGEN: A benchmarking library for retrieval-augmented generation. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 7640–7663, Miami, Florida, USA. Association for Computational Linguistics.

Rolinger, S. et al. (2025). Graph-of-thoughts for fact-checking with large language models. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 266–273, Vienna, Austria. Association for Computational Linguistics.

Schlichtkrull, M. et al. (2023). Averitec: A dataset for real-world claim verification with evidence from the web.

Shah, A. P. et al. (2026). Small language models: Architecture, evolution, and the future of artificial intelligence.

Thorne, J. et al. (2018). FEVER: a large-scale dataset for fact extraction and VERification. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 809–819, New Orleans, Louisiana. Association for Computational Linguistics.

Trivedi, H. et al. (2023). Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10014–10037, Toronto, Canada. Association for Computational Linguistics.

Wang, X. et al. (2024). Searching for best practices in retrieval-augmented generation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17716–17736, Miami, Florida, USA. Association for Computational Linguistics.

Willard, B. T. et al. (2023). Efficient guided generation for large language models.

Yang, A. et al. (2025). Qwen3 technical report. arXiv preprint arXiv:2505.09388.

Yao, S. et al. (2022). React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629.

Yoon, Y. et al. (2024). Hero at averitec: The herd of open large language models for verifying real-world claims. In Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER), pages 130–136.

Yoon, Y. et al. (2025). Team HUMANE at AVeriTeC 2025: HerO 2 for efficient fact verification. In Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER), pages 224–228, Vienna, Austria. Association for Computational Linguistics.

Yu, T. et al. (2024). In defense of rag in the era of long-context language models. arXiv preprint arXiv:2409.01666.