Research map / Large language models / Agents, code and math
Theory-Grounded Evaluation of Human-Like Fallacy Patterns in LLM Reasoning
Andrew Keenan Richardson, Ryan Othniel Kearns, Sean Moss, Vincent Wang, Philipp Koralus
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase model strength (higher Chatbot Arena Elo) | 38 LLMs · ETR fallacy problems | fallacy share of incorrect answers: r = 0.364, p = 0.0247 | Improved |
| increase model strength (Elo), exponential fit | 38 LLMs · ETR fallacy problems | fallacy share of incorrect answers: r = 0.407, p = 0.0113 | Improved |
| increase model strength (Elo), Spearman | 38 LLMs · ETR fallacy problems | fallacy share of incorrect answers: rho = 0.360, p = 0.0265 | Improved |
| increase model strength (higher Elo) | 38 LLMs · ETR fallacy problems | logical correctness: r = 0.004, p = 0.981 | No clear effect |
| increase model strength (higher Elo) | 38 LLMs · ETR fallacy problems | absolute proportion of fallacies | No clear effect |
| reverse premise order | claude-3.5-sonnet · ETR fallacy problems | fallacies blocked: 65.08% blocked (z = 4.74, p = 2.09e-06) | Improved |
| reverse premise order | gpt-3.5-turbo-1106 · ETR fallacy problems | fallacies blocked: 88.46% blocked (z = 4.36, p = 1.33e-05) | Improved |
| reverse premise order | claude-3.7-sonnet · ETR fallacy problems | fallacies blocked: 50.00% blocked (z = 4.23, p = 2.35e-05) | Improved |
| reverse premise order | o1-mini · ETR fallacy problems | fallacies blocked: 50.00% blocked (z = 0.58, p = 5.63e-01) | No clear effect |
| reverse premise order | gemma-3-27b-it · ETR fallacy problems | fallacies blocked: 20.39% blocked (z = 1.77, p = 7.60e-02) | No clear effect |
| reverse premise order | llama-3.2-3b-instruct · ETR fallacy problems | fallacies blocked: 20.00% blocked (z = 0.77, p = 4.41e-01) | No clear effect |
| reverse premise order | mistral-7b-instruct-v0.1 · ETR fallacy problems | fallacies blocked: 66.67% blocked (z = 1.42, p = 1.55e-01) | No clear effect |
Closest papers by meaning
- RULEBREAKERS: Challenging LLMs at the Crossroads between Formal Logic and Human-like Reasoning (ICML 2025 · similarity 0.91)
- An Enigma of Artificial Reason: Investigating the Production-Evaluation Gap in Large Reasoning Models (NeurIPS 2026 · similarity 0.89)
- Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement (ICLR 2024 · similarity 0.89)
- The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity (NeurIPS 2025 · similarity 0.88)
- Are Language Models Efficient Reasoners? A Perspective from Logic Programming (NeurIPS 2025 · similarity 0.88)
- Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens (NeurIPS 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
