Research map / Large language models / Retrieval and knowledge
RaCoT: Plug-and-Play Contrastive Example Generation Mechanism for Enhanced LLM Reasoning Reliability
Kaitong Cai, Jusheng Zhang, Yijia Fan, Jing Yang, Keze Wang
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 |
|---|---|---|---|
| add contrast prompting (ablation) | RaCoT · PopQA | accuracy: 55.1 -> 59.9 | Improved |
| add contrast prompting (ablation) | RaCoT · TriviaQA-unfiltered | accuracy: 69.6 -> 73.8 | Improved |
| add post-retrieval ranking (ablation) | RaCoT · PopQA | accuracy: 57.9 -> 59.9 | Improved |
| add similarity filtering for contrast question (ablation) | RaCoT · TriviaQA-unfiltered | accuracy: 72.5 -> 73.8 | Improved |
| use weaker teacher (GPT-4o-mini) (ablation) | RaCoT · PopQA | accuracy: 59.9 -> 58.5 | No clear effect |
| use RaCoT instead of RankRAG | Qwen2.5-7B · PopQA | accuracy: 66.4 -> 68.3 | Improved |
| use RaCoT instead of RankRAG | LLaMA3-8B · PopQA | accuracy: 58.5 -> 59.9 | Improved |
| use RaCoT instead of RankRAG | Qwen2.5-7B · OpenBookQA | accuracy: 87.5 -> 88.2 | No clear effect |
| distractor injection | RaCoT · PopQA and ARC-Challenge | accuracy drop: drop of 8.6% vs 16.5% RankRAG, 15.3% IterDRAG | Improved |
| distractor injection | RaCoT · PopQA (retrieval distractors) | accuracy drop: -8.7% vs -19.6% for Self-RAG/RQ-RAG | Improved |
Closest papers by meaning
- Retrieval is Not Enough: Enhancing RAG through Test-Time Critique and Optimization (NeurIPS 2025 · similarity 0.88)
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (ICLR 2024 · similarity 0.88)
- Making Retrieval-Augmented Language Models Robust to Irrelevant Context (ICLR 2024 · similarity 0.87)
- Thinking on the Fly: Test-Time Reasoning Enhancement via Latent Thought Policy Optimization (ICLR 2026 · similarity 0.87)
- Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts (ICLR 2026 · similarity 0.87)
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 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.
