Research map / Large language models / Agents, code and math
CodeIPPrompt: Intellectual Property Infringement Assessment of Code Language Models
Zhiyuan Yu, Yuhao Wu, Ning Zhang, Chenguang Wang, Yevgeniy Vorobeychik, Chaowei Xiao
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 |
|---|---|---|---|
| Dynamic token filtering (k=10, roll back when similarity >0.5) (ablation) | CodeParrot-110M and CodeParrot-1.5B · GitHub-sourced prompts / HumanEval | EM and pass@1: EM drops from 0.98 to 0.46 and 0.99 to 0.40; pass@1 drops from 3.41% to 1.17% and 3.84% to 1.46% | Mixed |
| Fine-tuning on ~17K public-domain code snippets (ablation) | CodeParrot and CodeRL models · GitHub-sourced prompts | EM: CodeParrot-110M EM 0.98 to 0.71; CodeParrot-1.5B 0.99 to 0.67 | No clear effect |
| Fine-tuning on public-domain code (ablation) | CodeT5-large · GitHub-sourced prompts | EM: EM rises from 0.11 to 0.71 | Hurt |
| Prompts derived from the model's own training dataset (vs GitHub prompts) | open-source models · training-dataset prompts, n=5 | EM and EP: higher similarity and probability generally | Improved |
| Increasing number of generations from 1 to 1000 | Copilot and Codex · GitHub-sourced prompts | EM and EP | Improved |
| Prompt length (longer prompts) | code LMs · GitHub-sourced prompts | similarity score: Pearson rho=0.0378, p=0.1472 | No clear effect |
| Commercial products (Codex, Copilot) vs open-source models | Codex and Copilot · GitHub-sourced prompts | EM: Codex 0.68 and Copilot 0.56 on permissive vs CodeParrot-110M 0.98 | Mixed |
Closest papers by meaning
- Evaluating Copyright Takedown Methods for Language Models (NeurIPS 2024 · similarity 0.85)
- Can LLMs Obfuscate Code? A Systematic Analysis of Large Language Models into Assembly Code Obfuscation (AAAI 2025 · similarity 0.84)
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code (ICLR 2025 · similarity 0.84)
- Automatically Testing Functional Properties of Code Translation Models (AAAI 2024 · similarity 0.84)
- EvoCodeBench: An Evolving Code Generation Benchmark with Domain-Specific Evaluations (NeurIPS 2024 · similarity 0.84)
- Evaluation of Large Language Models on Code Obfuscation (Student Abstract) (AAAI 2024 · similarity 0.84)
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.
