Research map / Large language models / Agents, code and math
The Curious Case of Analogies: Investigating Analogical Reasoning in Large Language Models
Taewhoo Lee, Minju Song, Chanwoong Yoon, Jungwoo Park, Jaewoo Kang
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 |
|---|---|---|---|
| replace the first entity pair of incorrect analogies with a correct pair (ablation) | Gemma-7B · Proportional analogies | corrected fraction: +38.4% | Improved |
| patch e2 hidden representation into the link position (ablation) | Gemma-7B · Proportional analogies (remaining errors) | corrected fraction: +38.1% | Improved |
| block attention from resolution token to e1 (ablation) | Qwen2.5-14B · Proportional analogies | accuracy | No clear effect |
| block attention from resolution token to e2 or e3 in mid-upper layers (ablation) | Qwen2.5-14B · Proportional analogies | accuracy | Hurt |
| compare relational vs attributive info decoded in correct vs incorrect cases (ablation) | Qwen2.5-14B · Proportional analogies | decoded information rate | Mixed |
| Mutual Alignment Score for target vs distractor stories | Qwen2.5-14B-Instruct · StoryAnalogy | relative MAS | Mixed |
Closest papers by meaning
- Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement (ICLR 2024 · similarity 0.87)
- Reasoning Scaffolding: Distilling the Flow of Thought from LLMs (ICLR 2026 · similarity 0.87)
- Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models (ICLR 2025 · similarity 0.87)
- Benchmarking and Understanding Compositional Relational Reasoning of LLMs (AAAI 2025 · similarity 0.87)
- Large Language Models Think Too Fast To Explore Effectively (NeurIPS 2025 · similarity 0.87)
- CogMath: Assessing LLMs' Authentic Mathematical Ability from a Human Cognitive Perspective (ICML 2025 · similarity 0.87)
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.
