Research map / Large language models / Reasoning and chain of thought
Evaluating and Improving Tool-Augmented Computation-Intensive Math Reasoning
Beichen Zhang, Kun Zhou, Xilin Wei, Xin Zhao, Jing Sha, Shijin Wang, Ji-Rong Wen
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 |
|---|---|---|---|
| DELI two-stage deliberation vs best baseline ReAct | DELI · CARP | accuracy (%): 64.11 -> 73.46 | Improved |
| DELI vs ReAct | DELI · Algebra | accuracy (%): 54.51 -> 59.65 | Improved |
| DELI vs ReAct | DELI · Counting & Probability | accuracy (%): 41.77 -> 39.03 | Hurt |
| DELI vs PHP | DELI · Number Theory | accuracy (%): 35.37 -> 33.15 | Hurt |
| add retrieved exemplars (Retrieval CoT vs Random CoT) | CoT prompting · CARP | accuracy (%): 49.39 -> 63.93 | Improved |
| add tool interfaces (ReAct vs Random CoT) | CoT prompting · CARP | accuracy (%): 49.39 -> 64.11 | Improved |
| ReAct-style interleaved reasoning vs PAL program generation | tool-augmented prompting · CARP | accuracy (%): 40.00 -> 64.11 | Improved |
| iterate without two-stage deliberation (Iterative CoT vs Random CoT/Retrieval CoT) (ablation) | Retrieval CoT · CARP | accuracy (%): 63.93 -> 61.27 | Hurt |
| iterate without two-stage deliberation (Iterative ReAct vs ReAct) (ablation) | ReAct · CARP | accuracy (%): 64.11 -> 61.17 | Hurt |
| add DELI deliberation on top of ReAct (ablation) | ReAct · Average of 7 datasets | accuracy (%): 48.07 -> 50.85 | Improved |
| add DELI on top of CoT initialization (ablation) | CoT prompting · CARP, Algebra, Counting & Probability | accuracy | Improved |
| increase maximum iteration turns for DELI (ablation) | DELI · CARP, Algebra, Prealgebra | accuracy | Improved |
Closest papers by meaning
- Premise-Augmented Reasoning Chains Improve Error Identification in Math reasoning with LLMs (ICML 2025 · similarity 0.92)
- DAG-Math: Graph-Guided Mathematical Reasoning in LLMs (ICLR 2026 · similarity 0.91)
- Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning (NeurIPS 2025 · similarity 0.91)
- To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning (ICLR 2025 · similarity 0.91)
- Mind the Gap: Bridging Thought Leap for Improved Chain-of-Thought Tuning (NeurIPS 2025 · similarity 0.90)
- MathFimer: Enhancing Mathematical Reasoning by Expanding Reasoning Steps through Fill-in-the-Middle Task (ICLR 2026 · similarity 0.90)
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.
