Research map / Large language models / Reasoning and chain of thought
RBF++: Quantifying and Optimizing Reasoning Boundaries Across Measurable and Unmeasurable Capabilities for Chain-of-Thought Reasoning
Qiguang Chen, Libo Qin, Jinhao Liu, Yue Liao, Jiaqi Wang, Jingxuan Zhou, Wanxiang Che
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 |
|---|---|---|---|
| use Tool Usage instead of plain CoT | GPT-3.5-Turbo · BigGSM | accuracy: 57.00 -> 71.64 | Improved |
| use Program-of-Thought instead of Tool Usage | GPT-3.5-Turbo · BigGSM | accuracy: 71.64 -> 78.25 | Improved |
| add MARP prompting to CoT | GPT-3.5-Turbo · BigGSM | accuracy: 57.00 -> 64.37 | Improved |
| add MARP prompting to PoT | GPT-3.5-Turbo · BigGSM | accuracy: 78.25 -> 80.55 | Improved |
| add Least-to-Most prompting to CoT | GPT-3.5-Turbo · BigGSM | accuracy: 57.00 -> 58.25 | No clear effect |
| add Complex-CoT to CoT | GPT-3.5-Turbo · BigGSM | accuracy: 57.00 -> 59.78 | No clear effect |
| add Least-to-Most prompting to CoT | GPT-4o · M3CoT Total | accuracy: 68.68 -> 65.88 | Hurt |
| add Complex-CoT to CoT | GPT-4o · M3CoT Total | accuracy: 68.68 -> 66.95 | Hurt |
| add MARP to CoT | GPT-4o · M3CoT Total | accuracy: 68.68 -> 67.82 | Hurt |
| add MARP++ (perception and knowledge boundaries) to MARP | GPT-4o · M3CoT Total | accuracy: 67.82 -> 73.77 | Improved |
| add MARP++ to CoT | GPT-4o · M3CoT Science Natural | accuracy: 74.07 -> 78.93 | Improved |
| add Least-to-Most prompting to CoT | GPT-4o · M3CoT Mathematics Geometry | accuracy: 28.75 -> 45.00 | Mixed |
Closest papers by meaning
- Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought (NeurIPS 2024 · similarity 0.96)
- The CoT Encyclopedia: Analyzing, Predicting, and Controlling how a Reasoning Model will Think (ICLR 2026 · similarity 0.92)
- KAM-CoT: Knowledge Augmented Multimodal Chain-of-Thoughts Reasoning (AAAI 2024 · similarity 0.91)
- MME-CoT: Benchmarking Chain-of-Thought in Large Multimodal Models for Reasoning Quality, Robustness, and Efficiency (ICML 2025 · similarity 0.91)
- FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning (ICLR 2026 · similarity 0.91)
- Understanding Chain-of-Thought in LLMs through Information Theory (ICML 2025 · similarity 0.91)
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.
