Research map / Large language models / Agents, code and math
ConvexBench: Can LLMs Recognize Convex Functions?
Yepeng Liu, Yu Huang, Yu-Xiang Wang, Yingbin Liang, Yuheng Bu
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 composition depth for one-shot reasoning (ablation) | frontier LLMs · ConvexBench | macro F1: 1.0 at depth 2 -> approximately 0.2 at depth 100 | Hurt |
| add external AST decomposition of the expression (ablation) | One-shot Reasoning · ConvexBench | macro F1 | Improved |
| add recursive agentic step-by-step reasoning on top of decomposition (ablation) | One-shot Reasoning w/ Decomp · ConvexBench | macro F1 | Improved |
| add dependency-focused context instead of full cumulative context (ablation) | Agentic Reasoning · ConvexBench | macro F1: F1 1.0 at depth 100 for GPT-5, Gemini-2.5-Pro, Qwen3-30B | Improved |
| use coarser decomposition (longer sub-functions, 10/50/100 chars) (ablation) | One-shot Reasoning w/ Decomp · ConvexBench | macro F1 | Hurt |
| use finer decomposition (sub-function length 10 vs 100) (ablation) | Agentic Reasoning · ConvexBench depth 100 | macro F1 / recursion steps: recursion steps 22 -> 146 | Improved |
| apply agentic reasoning to frontier models already near ceiling with decomposition (ablation) | GPT-5, Gemini-2.5-Pro · ConvexBench | macro F1 / cost: up to 10x inference cost | Mixed |
Closest papers by meaning
- Putnam-AXIOM: A Functional & Static Benchmark for Measuring Higher Level Mathematical Reasoning in LLMs (ICML 2025 · similarity 0.87)
- CRANE: Reasoning with constrained LLM generation (ICML 2025 · similarity 0.86)
- APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning (NeurIPS 2025 · similarity 0.86)
- Solving Inequality Proofs with Large Language Models (NeurIPS 2025 · similarity 0.86)
- Benchmarking Abstract and Reasoning Abilities Through A Theoretical Perspective (ICML 2025 · similarity 0.86)
- Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus (NeurIPS 2024 · similarity 0.86)
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.
