Research map / Large language models / Language, safety and interpretability
Rethinking Code Complexity Through the Lens of Large Language Models
Chen Xie, Xiaodong Gu, Yuling Shi, Beijun Shen
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 |
|---|---|---|---|
| rewrite code to lower LM-CC while holding cyclomatic complexity fixed | DeepSeek-V3 · Program repair | pass@1: 13.4% -> 16.2% | Improved |
| rewrite code to lower LM-CC while holding cyclomatic complexity fixed | DeepSeek-V3 · Code translation | pass@1: 42.2% -> 46.5% | Improved |
| rewrite code to lower LM-CC while holding cyclomatic complexity fixed | DeepSeek-V3 · Execution reasoning | pass@1: 94.1% -> 95.0% | Improved |
| control for code length in HC correlation | Halstead Complexity · Program repair | Spearman correlation: -0.97 -> -0.93 | Hurt |
| control for code length in HC correlation | Halstead Complexity · Code translation | Spearman correlation: -0.90 -> -0.70 | Hurt |
| control for code length in CC correlation | Cyclomatic Complexity · Execution reasoning | Spearman correlation: -0.78 -> -0.67 | Hurt |
| combine TotalBranch and TotalCompLevel into LM-CC (ablation) | LM-CC · 16 task-model combinations | significant correlations: 14/16 -> 16/16 | Improved |
| add entropy-based boundaries to syntax-only decomposition (ablation) | LM-CC · Program repair, DeepSeek-V3 | partial Spearman correlation: -0.88 -> -0.93 | Improved |
| add entropy-based boundaries to syntax-only decomposition (ablation) | LM-CC · Code translation, DeepSeek-V3 | partial Spearman correlation: -0.78 -> -0.97 | Improved |
| add syntactic delimiters to entropy-only decomposition (ablation) | LM-CC · Program repair, DeepSeek-V3 | partial Spearman correlation: -0.72 -> -0.93 | Improved |
| add syntactic delimiters to entropy-only decomposition (ablation) | LM-CC · Program repair, GPT-4o-mini | partial Spearman correlation: -0.93 -> -0.85 | Hurt |
| add syntactic delimiters to entropy-only decomposition (ablation) | LM-CC · Execution reasoning, DeepSeek-V3 | partial Spearman correlation: - -> -0.92 | Improved |
Closest papers by meaning
- CodeMMLU: A Multi-Task Benchmark for Assessing Code Understanding & Reasoning Capabilities of CodeLLMs (ICLR 2025 · similarity 0.87)
- How efficient is LLM-generated code? A rigorous & high-standard benchmark (ICLR 2025 · similarity 0.87)
- Unsupervised Evaluation of Code LLMs with Round-Trip Correctness (ICML 2024 · similarity 0.87)
- CodeCrash: Exposing LLM Fragility to Misleading Natural Language in Code Reasoning (NeurIPS 2025 · similarity 0.86)
- CoRe: Benchmarking LLMs’ Code Reasoning Capabilities through Static Analysis Tasks (NeurIPS 2025 · similarity 0.86)
- EffiCoder: Enhancing Code Generation in Large Language Models through Efficiency-Aware Fine-tuning (ICML 2025 · similarity 0.85)
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.
