Research map / Label-efficient and robust learning / Recognition and long-tail
CodeCMR: Cross-Modal Retrieval For Function-Level Binary Source Code Matching
Zeping Yu, Wenxin Zheng, Jiaqi Wang, Qiyi Tang, Sen Nie, Shi Wu
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 |
|---|---|---|---|
| replace random negative sampling with norm weighted sampling (s=5) (ablation) | CodeCMR · gcc-x64-O0 Binary2Source | recall@1: 81.9 -> 90.2 | Improved |
| replace distance weighted sampling with norm weighted sampling (s=5) (ablation) | CodeCMR · clang-arm-O3 Binary2Source | recall@1: 77.4 -> 87.3 | Improved |
| use small s=0.5 instead of s=1 (distance weighted) (ablation) | norm weighted sampling · gcc-x64-O0 Binary2Source | recall@1: 86.2 -> 85.3 | Hurt |
| increase s from 5 to 8 or 10 (ablation) | norm weighted sampling · gcc-x64-O0 | recall@1 | Hurt |
| replace pre-trained BERT node embeddings with end-to-end HBMP (ablation) | DPCNN binary-source model · gcc-x64-O0 Binary2Source | recall@1: 74.3 -> 80.8 | Improved |
| use DPCNN instead of LSTM for source code encoding (ablation) | HBMP binary encoder · gcc-x64-O0 Binary2Source | recall@1: 63.7 -> 80.8 | Improved |
| add integer number feature (Integer-LSTM vs LSTM) (ablation) | integer literal encoder · gcc-x64-O0 Binary2Source | recall@1: 10.7 -> 12.3 | Improved |
| replace Hungarian algorithm with Hier-LSTM on strings (ablation) | string literal matching · gcc-x64-O0 Binary2Source | recall@1: 33.9 -> 42.4 | Improved |
| add code literal features (strings, integers) to DPCNN+HBMP semantic features (ablation) | CodeCMR · gcc-x64-O0 Binary2Source | recall@1: 80.8 -> 81.9 | Improved |
Closest papers by meaning
- Source Code Foundation Models are Transferable Binary Analysis Knowledge Bases (NeurIPS 2024 · similarity 0.87)
- Is Function Similarity Over-Engineered? Building a Benchmark (NeurIPS 2024 · similarity 0.84)
- Towards A Generalist Code Embedding Model Based On Massive Data Synthesis (NeurIPS 2025 · similarity 0.83)
- CodeMMR: Bridging Natural Language, Code, and Image for Unified Retrieval (CVPR 2026 · similarity 0.82)
- PinNet: Pinpoint Instructive Information for Retrieval Augmented Code-to-Text Generation (ICML 2024 · similarity 0.82)
- Decompile-Bench: Million-Scale Binary-Source Function Pairs for Real-World Binary Decompilation (NeurIPS 2025 · similarity 0.82)
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.
