pith. sign in

arxiv: 2606.31039 · v1 · pith:63FPT3RXnew · submitted 2026-06-30 · 💻 cs.CL

Truth or Sophistry? LoFa: A Benchmark for LLM Robustness Against Logical Fallacies

Xudong Shen , Li Yuan , Ye Chen , Xin Wu , Yi Cai , Zhiyong Wu This is my paper

Pith reviewed 2026-07-01 06:09 UTC · model grok-4.3

classification 💻 cs.CL
keywords logical fallaciesLLM robustnessbenchmark constructionmulti-agent pipelineLFR@k metricadversarial persuasionvulnerability profilesdebate evaluation
0
0 comments X

The pith

The LoFa benchmark shows LLMs display distinct vulnerability profiles to different logical fallacies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces LoFa to test how well large language models resist persuasion by logical fallacies rather than just detecting them. It builds the benchmark with a multi-agent pipeline that creates pairs of factual questions and fallacious arguments, then runs them through a multi-round debate setup. The LFR@k metric isolates fallacy resistance from gaps in the model's knowledge. Experiments find that robustness levels differ across fallacy types and produce unique weakness patterns for each model tested.

Core claim

LoFa is built through a multi-agent pipeline that pairs factual questions with fallacious arguments and is evaluated in a multi-round debate framework. The Logical Fallacy Resistance at k (LFR@k) metric quantifies resistance to fallacious attacks while separating it from inherent knowledge limitations. Experiments show that LLMs exhibit varying levels of robustness across different fallacy types, revealing distinct vulnerability profiles among models.

What carries the argument

The multi-agent pipeline that generates pairs of factual questions and fallacious arguments, together with the Logical Fallacy Resistance at k (LFR@k) metric that measures sustained resistance in debate.

If this is right

  • Models can be profiled for targeted weaknesses against specific fallacy types.
  • Standardized tests allow direct comparison of robustness between different LLMs.
  • Multi-round debate evaluation reveals resilience limits not visible in single-turn tests.
  • The LFR@k metric enables focused measurement of fallacy resistance independent of factual knowledge.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The construction method could extend to testing resistance against other manipulative patterns such as emotional appeals or selective framing.
  • If adopted widely, the benchmark might inform safety evaluations for LLMs used in debate-heavy domains like law or public policy.
  • Observed differences across models suggest that training data composition influences which fallacies are harder to resist.

Load-bearing premise

The multi-agent pipeline produces fallacious arguments that are both logically invalid and persuasive enough to serve as valid tests without construction artifacts that would confound the robustness measurement.

What would settle it

Human judges rating the generated arguments as non-fallacious or non-persuasive, or models showing identical performance across fallacy types when tested on the same questions without the pipeline.

Figures

Figures reproduced from arXiv: 2606.31039 by Li Yuan, Xin Wu, Xudong Shen, Ye Chen, Yi Cai, Zhiyong Wu.

Figure 1
Figure 1. Figure 1: Overview of the multi-stage evaluation frame [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: From Ambiguous Opinion to Measurable Error: The necessity of objective ground truth for fallacy evaluation. model resilience under sustained attacks, depicted in [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the multi-agent pipeline for constructing the LoFa dataset with iterative quality assurance. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: LFR@k (1,2,3) of the LLMs for 10 logical fallacies on the NQ1 dataset. 5.2 Experimental Results [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The Architectural Fingerprint in Fallacy Resis [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The heatmap visualizes the top-2 projected tokens across transformer layers for Llama-3.1-8b-instruct [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Performance of Direct vs. CoT Prompting on [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Results on the NQ1 dataset SM Eq AA AH RH HG SS FD FC CR 47.8 66.5 37.0 31.4 95.4 83.5 88.4 69.8 88.2 69.2 92.2 76.6 91.4 77.1 81.9 57.7 87.4 73.5 93.2 81.8 SM Eq AA AH RH HG SS FD FC CR 64.0 90.3 94.8 67.1 93.2 84.0 86.4 55.6 98.2 79.3 95.882.5 94.0 83.1 96.1 79.3 97.0 73.1 97.0 85.1 llama-3.1-8B llama-3.1-70B llama-3.1-405B gpt-3.5-turbo gpt-4 deepseek-v3 [PITH_FULL_IMAGE:figures/full_fig_p032_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Results on the TruthfulQA dataset SM Eq AA AH RH HG SS FD FC CR 39.051.1 32.1 16.6 87.7 78.1 70.1 88.3 54.9 69.2 84.2 65.0 85.1 72.0 70.7 54.3 80.9 70.5 93.6 76.6 SM Eq AA AH RH HG SS FD FC CR 51.1 91.0 87.3 24.4 93.3 78.7 79.1 46.0 97.0 71.3 91.3 67.5 96.2 74.4 88.4 62.5 94.0 62.5 93.6 74.4 llama-3.1-8B llama-3.1-70B llama-3.1-405B gpt-3.5-turbo gpt-4 deepseek-v3 [PITH_FULL_IMAGE:figures/full_fig_p032_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Results on the Boolq dataset SM Eq AA AH RH HG SS FD FC CR 56.364.2 37.0 25.2 94.5 89.0 80.9 67.3 89.4 76.1 92.980.0 97.591.0 89.3 75.6 92.0 78.7 95.6 92.1 SM Eq AA AH RH HG SS FD FC CR 64.2 91.1 95.0 37.0 94.7 82.3 88.2 37.3 96.5 84.7 96.291.1 97.5 87.0 94.2 78.2 97.7 85.7 95.6 87.1 llama-3.1-8B llama-3.1-70B llama-3.1-405B gpt-3.5-turbo gpt-4 deepseek-v3 [PITH_FULL_IMAGE:figures/full_fig_p032_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Results on the NQ2 dataset G.2 Visual Analysis of Successful Fallacy Resistance To validate our findings regarding the Cognitive Void, we extend our layer-wise analysis to instances where Llama-3.1-8b-instruct successfully resists fallacious attacks [PITH_FULL_IMAGE:figures/full_fig_p032_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Internal Dynamics of Successful Resistance. (1) Clean Context: The model exhibits early and decisive convergence to the correct token “James” (Layer 23) with high confidence (dark blue). (2) Fallacious Context (Resistance): The model successfully outputs the correct answer but exhibits a delayed and hesitant trajectory. Note the extended sequence of “none” tokens (Layers 20–26). Crucially, unlike failure … view at source ↗
read the original abstract

Large Language Models (LLMs) exhibit strong semantic capabilities, yet their resilience to manipulative linguistic patterns such as logical fallacies remains underexplored. Prior work has primarily examined whether LLMs can identify or classify fallacies, leaving their robustness against fallacious persuasion insufficiently studied. To address this gap, we introduce LoFa (Logical Fallacy), a comprehensive benchmark for evaluating LLM robustness against fallacies. LoFa is constructed through a multi-agent pipeline that pairs factual questions with fallacious arguments, and is accompanied by a multi-round debate framework for assessing model resilience under sustained adversarial persuasion. To disentangle fallacy robustness from a model's inherent knowledge limitations, we further propose Logical Fallacy Resistance at k (LFR@k), a metric that quantifies resistance to fallacious attacks. Experiments show that LLMs exhibit varying levels of robustness across different fallacy types, revealing distinct vulnerability profiles among models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces LoFa, a benchmark for LLM robustness to logical fallacies. It is built via a multi-agent pipeline pairing factual questions with fallacious arguments, evaluated in a multi-round debate framework, and accompanied by the LFR@k metric that aims to isolate fallacy resistance from knowledge limitations. Experiments are claimed to demonstrate varying robustness across fallacy types and distinct vulnerability profiles among models.

Significance. If the generated fallacies are shown to be valid (logically invalid yet persuasive without construction artifacts) and the experimental results are reproducible, the benchmark and metric could fill a gap left by prior fallacy-classification work and support more targeted robustness improvements in LLMs.

major comments (2)
  1. [Abstract] Abstract: the central claim that 'Experiments show that LLMs exhibit varying levels of robustness across different fallacy types' is unsupported by any quantitative results, tables, error analysis, or validation of the generated fallacies, leaving the experimental contribution unassessable.
  2. [Benchmark Construction] Benchmark Construction (multi-agent pipeline): no validation procedure (human judgment, logical entailment checks, or inter-annotator agreement) is described to confirm that the generated arguments are both fallacious and free of confounding artifacts; this directly affects the validity of the LFR@k measurements and the reported vulnerability profiles.
minor comments (2)
  1. [Metric Definition] The LFR@k definition would benefit from an explicit equation or pseudocode to clarify how 'resistance at k' is computed across debate rounds.
  2. [Results] Figure or table captions for model comparisons should include the exact fallacy types and k values used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments point-by-point below. Where the concerns are valid, we commit to revisions that strengthen the manuscript without altering its core contributions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that 'Experiments show that LLMs exhibit varying levels of robustness across different fallacy types' is unsupported by any quantitative results, tables, error analysis, or validation of the generated fallacies, leaving the experimental contribution unassessable.

    Authors: The abstract is a high-level summary; the full manuscript contains the Experiments section with quantitative LFR@k tables across fallacy types and models, plus vulnerability profile analysis. However, we agree the abstract could better signal the empirical support. We will revise the abstract to briefly reference key quantitative outcomes (e.g., average LFR@k ranges) while preserving its length. revision: yes

  2. Referee: [Benchmark Construction] Benchmark Construction (multi-agent pipeline): no validation procedure (human judgment, logical entailment checks, or inter-annotator agreement) is described to confirm that the generated arguments are both fallacious and free of confounding artifacts; this directly affects the validity of the LFR@k measurements and the reported vulnerability profiles.

    Authors: This is a fair observation. The current manuscript relies on the multi-agent generation process without explicit post-generation validation. We will add a dedicated validation subsection describing human evaluation (with inter-annotator agreement) and logical entailment checks on a sample of generated pairs to confirm fallaciousness and absence of artifacts. This will directly support the reliability of LFR@k and the reported profiles. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces the LoFa benchmark via a multi-agent pipeline and defines the LFR@k metric directly from its outputs. No equations, fitted parameters renamed as predictions, self-citations, or uniqueness claims appear in the provided text. The central contribution is an empirical benchmark and evaluation framework whose construction does not reduce to its own inputs by definition. This is the standard non-circular case for benchmark papers.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no identifiable free parameters, axioms, or invented entities; all details of the pipeline and metric definition are absent.

pith-pipeline@v0.9.1-grok · 5694 in / 908 out tokens · 21406 ms · 2026-07-01T06:09:55.552298+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 21 canonical work pages · 3 internal anchors

  1. [1]

    Aho and Jeffrey D

    Alfred V. Aho and Jeffrey D. Ullman , title =. 1972

    1972

  2. [2]

    Publications Manual , year = "1983", publisher =

    1983

  3. [3]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243

    work page doi:10.1145/322234.322243 1981

  4. [4]

    Scalable training of

    Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of

  5. [5]

    Dan Gusfield , title =. 1997

    1997

  6. [6]

    Tetreault , title =

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url =

    2015

  7. [7]

    A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =

    Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages =

  8. [8]

    2025 , eprint=

    Empowering LLMs with Logical Reasoning: A Comprehensive Survey , author=. 2025 , eprint=

    2025

  9. [9]

    Problems of Education in the 21st Century , volume=

    Critical thinking in education: why to avoid logical fallacies? , author=. Problems of Education in the 21st Century , volume=. 2014 , publisher=

    2014

  10. [10]

    On-line Article (preprint), doi , volume=

    Logical fallacies , author=. On-line Article (preprint), doi , volume=

  11. [11]

    2025 , eprint=

    A Survey of Large Language Models , author=. 2025 , eprint=

    2025

  12. [12]

    2024 , eprint=

    Large Language Models Meet NLP: A Survey , author=. 2024 , eprint=

    2024

  13. [13]

    Nature communications , volume=

    Benchmarking large language models for biomedical natural language processing applications and recommendations , author=. Nature communications , volume=. 2025 , publisher=

    2025

  14. [14]

    2012 , publisher=

    Logically fallacious: the ultimate collection of over 300 logical fallacies (Academic Edition) , author=. 2012 , publisher=

    2012

  15. [15]

    arXiv preprint arXiv:2202.13758 , year=

    Logical fallacy detection , author=. arXiv preprint arXiv:2202.13758 , year=

    work page arXiv

  16. [16]

    arXiv preprint arXiv:2312.09085 , year=

    The earth is flat because...: Investigating llms' belief towards misinformation via persuasive conversation , author=. arXiv preprint arXiv:2312.09085 , year=

    work page arXiv

  17. [17]

    B ool Q : Exploring the Surprising Difficulty of Natural Yes/No Questions

    Clark, Christopher and Lee, Kenton and Chang, Ming-Wei and Kwiatkowski, Tom and Collins, Michael and Toutanova, Kristina. B ool Q : Exploring the Surprising Difficulty of Natural Yes/No Questions. Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long ...

    work page doi:10.18653/v1/n19-1300 2019

  18. [18]

    Transactions of the Association for Computational Linguistics , volume=

    Natural questions: a benchmark for question answering research , author=. Transactions of the Association for Computational Linguistics , volume=. 2019 , publisher=

    2019

  19. [19]

    T ruthful QA : Measuring How Models Mimic Human Falsehoods

    Lin, Stephanie and Hilton, Jacob and Evans, Owain. T ruthful QA : Measuring How Models Mimic Human Falsehoods. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022. doi:10.18653/v1/2022.acl-long.229

    work page doi:10.18653/v1/2022.acl-long.229 2022

  20. [20]

    arXiv preprint arXiv:2406.07400 , year=

    Guiding LLM Temporal Logic Generation with Explicit Separation of Data and Control , author=. arXiv preprint arXiv:2406.07400 , year=

    work page arXiv

  21. [21]

    DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning , author=. arXiv preprint arXiv:2501.12948 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  22. [22]

    PeerJ Computer Science , volume=

    LOGIC: LLM-originated guidance for internal cognitive improvement of small language models in stance detection , author=. PeerJ Computer Science , volume=. 2024 , publisher=

    2024

  23. [23]

    arXiv preprint arXiv:1707.06002 , year=

    Argotario: Computational argumentation meets serious games , author=. arXiv preprint arXiv:1707.06002 , year=

    work page arXiv

  24. [24]

    L ogic B ench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models

    Parmar, Mihir and Patel, Nisarg and Varshney, Neeraj and Nakamura, Mutsumi and Luo, Man and Mashetty, Santosh and Mitra, Arindam and Baral, Chitta. L ogic B ench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Paper...

    work page doi:10.18653/v1/2024.acl-long.739 2024

  25. [25]

    2023 , eprint=

    Universal and Transferable Adversarial Attacks on Aligned Language Models , author=. 2023 , eprint=

    2023

  26. [26]

    Proceedings of the 37th International Conference on Neural Information Processing Systems , articleno =

    Wei, Alexander and Haghtalab, Nika and Steinhardt, Jacob , title =. Proceedings of the 37th International Conference on Neural Information Processing Systems , articleno =. 2023 , publisher =

    2023

  27. [27]

    DeepSeek-V3 Technical Report

    Deepseek-v3 technical report , author=. arXiv preprint arXiv:2412.19437 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  28. [28]

    GPT-4 Technical Report

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  29. [29]

    Advances in neural information processing systems , volume=

    Language models are few-shot learners , author=. Advances in neural information processing systems , volume=

  30. [30]

    arXiv e-prints , pages=

    The llama 3 herd of models , author=. arXiv e-prints , pages=

  31. [31]

    2025 , eprint=

    Logical Reasoning in Large Language Models: A Survey , author=. 2025 , eprint=

    2025

  32. [32]

    arXiv preprint arXiv:2504.20213 , year=

    Can large language models learn formal logic? a data-driven training and evaluation framework , author=. arXiv preprint arXiv:2504.20213 , year=

    work page arXiv

  33. [33]

    2024 , eprint=

    Flee the Flaw: Annotating the Underlying Logic of Fallacious Arguments Through Templates and Slot-filling , author=. 2024 , eprint=

    2024

  34. [34]

    2024 4th International Conference on Artificial Intelligence, Robotics, and Communication (ICAIRC) , pages=

    Robustness of large language models against adversarial attacks , author=. 2024 4th International Conference on Artificial Intelligence, Robotics, and Communication (ICAIRC) , pages=. 2024 , organization=

    2024

  35. [35]

    2025 , eprint=

    Evaluating LLMs Robustness in Less Resourced Languages with Proxy Models , author=. 2025 , eprint=

    2025

  36. [36]

    Contemporary Educational Psychology , volume=

    Individual differences in the analysis of informal reasoning fallacies , author=. Contemporary Educational Psychology , volume=. 2007 , publisher=

    2007

  37. [37]

    Social Sciences , volume=

    The relationship between cognitive bias and logical fallacies in egyptian society , author=. Social Sciences , volume=

  38. [38]

    Knowledge-Based Systems , volume=

    Robust and explainable identification of logical fallacies in natural language arguments , author=. Knowledge-Based Systems , volume=. 2023 , publisher=

    2023

  39. [39]

    A Logical Fallacy-Informed Framework for Argument Generation

    Mouchel, Luca and Paul, Debjit and Cui, Shaobo and West, Robert and Bosselut, Antoine and Faltings, Boi. A Logical Fallacy-Informed Framework for Argument Generation. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025. doi:10....

    work page doi:10.18653/v1/2025.naacl-long.374 2025

  40. [40]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    MAPS: Multi-Agent Personality Shaping for Collaborative Reasoning , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2026 , month=. doi:10.1609/aaai.v40i19.38669 , number=

    work page doi:10.1609/aaai.v40i19.38669 2026

  41. [41]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    MARS: Multi-Agent Adaptive Reasoning with Socratic Guidance for Automated Prompt Optimization , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2026 , month=. doi:10.1609/aaai.v40i19.38668 , number=

    work page doi:10.1609/aaai.v40i19.38668 2026

  42. [42]

    CODEMENV : Benchmarking Large Language Models on Code Migration

    Cheng, Keyuan and Shen, Xudong and Yang, Yihao and Wang, Tengyue and Cao, Yang and Ali, Muhammad Asif and Wang, Hanbin and Hu, Lijie and Wang, Di. CODEMENV : Benchmarking Large Language Models on Code Migration. Findings of the Association for Computational Linguistics: ACL 2025. 2025. doi:10.18653/v1/2025.findings-acl.140

    work page doi:10.18653/v1/2025.findings-acl.140 2025

  43. [43]

    Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence , pages=

    Collaborative multi-LoRA experts with achievement-based multi-tasks loss for unified multimodal information extraction , author=. Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence , pages=

  44. [44]

    The Earth is Flat because...: Investigating LLM s' Belief towards Misinformation via Persuasive Conversation

    Xu, Rongwu and Lin, Brian and Yang, Shujian and Zhang, Tianqi and Shi, Weiyan and Zhang, Tianwei and Fang, Zhixuan and Xu, Wei and Qiu, Han. The Earth is Flat because...: Investigating LLM s' Belief towards Misinformation via Persuasive Conversation. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pa...

    work page doi:10.18653/v1/2024.acl-long.858 2024

  45. [45]

    Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence,

    Empowering LLMs with Logical Reasoning: A Comprehensive Survey , author =. Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence,. 2025 , month =. doi:10.24963/ijcai.2025/1155 , url =

    work page doi:10.24963/ijcai.2025/1155 2025

  46. [46]

    A rgotario: Computational Argumentation Meets Serious Games

    Habernal, Ivan and Hannemann, Raffael and Pollak, Christian and Klamm, Christopher and Pauli, Patrick and Gurevych, Iryna. A rgotario: Computational Argumentation Meets Serious Games. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2017. doi:10.18653/v1/D17-2002

    work page doi:10.18653/v1/d17-2002 2017

  47. [47]

    How Susceptible Are LLM s to Logical Fallacies?

    Payandeh, Amirreza and Pluth, Dan and Hosier, Jordan and Xiao, Xuesu and Gurbani, Vijay K. How Susceptible Are LLM s to Logical Fallacies?. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). 2024

    2024

  48. [48]

    MAFALDA : A Benchmark and Comprehensive Study of Fallacy Detection and Classification

    Helwe, Chadi and Calamai, Tom and Paris, Pierre-Henri and Clavel, Chlo \'e and Suchanek, Fabian. MAFALDA : A Benchmark and Comprehensive Study of Fallacy Detection and Classification. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024....

    work page doi:10.18653/v1/2024.naacl-long.270 2024

  49. [49]

    FA ct S core: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation

    Min, Sewon and Krishna, Kalpesh and Lyu, Xinxi and Lewis, Mike and Yih, Wen-tau and Koh, Pang and Iyyer, Mohit and Zettlemoyer, Luke and Hajishirzi, Hannaneh. FA ct S core: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023. doi:10.1...

    work page doi:10.18653/v1/2023.emnlp-main.741 2023

  50. [50]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Hybrid-DMKG: A Hybrid Reasoning Framework over Dynamic Multimodal Knowledge Graphs for Multimodal Multihop QA with Knowledge Editing , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=