Research map / Large language models / Language, safety and interpretability
Associating Natural Language Comment and Source Code Entities
Sheena Panthaplackel, Milos Gligoric, Raymond J. Mooney, Junyi Jessy Li
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 |
|---|---|---|---|
| Use the binary classifier instead of the return-line presence baseline | Comment-code entity association · Annotated test set | F1: 0.485 -> 0.610 | Improved |
| Use the CRF instead of the return-line presence baseline | Comment-code entity association · Annotated test set | F1: 0.485 -> 0.559 | Improved |
| Add 100 deleted examples to training | Binary classifier · Annotated test set | F1: 0.610 -> 0.603 | Hurt |
| Add 200 deleted examples to training | Binary classifier · Annotated test set | F1: 0.610 -> 0.614 | Improved |
| Add 500 deleted examples to training | Binary classifier · Annotated test set | F1: 0.610 -> 0.655 | Improved |
| Add 867 deleted examples to training | Binary classifier · Annotated test set | F1: 0.610 -> 0.677 | Improved |
| Add 100 deleted examples to training | CRF · Annotated test set | F1: 0.559 -> 0.582 | Improved |
| Add 200 deleted examples to training | CRF · Annotated test set | F1: 0.559 -> 0.585 | Improved |
| Add 500 deleted examples to training | CRF · Annotated test set | F1: 0.559 -> 0.599 | Improved |
| Add 867 deleted examples to training | CRF · Annotated test set | F1: 0.559 -> 0.618 | Improved |
| Add code embeddings (ablation) | Binary classifier · Annotated test set | F1: 0.562 -> 0.610 | Improved |
| Add comment embeddings (ablation) | Binary classifier · Annotated test set | F1: 0.587 -> 0.610 | Improved |
Closest papers by meaning
- Deep Just-In-Time Inconsistency Detection Between Comments and Source Code (AAAI 2021 · similarity 0.88)
- CoRNStack: High-Quality Contrastive Data for Better Code Retrieval and Reranking (ICLR 2025 · similarity 0.83)
- LILO: Learning Interpretable Libraries by Compressing and Documenting Code (ICLR 2024 · similarity 0.82)
- ObscuraCoder: Powering Efficient Code LM Pre-Training Via Obfuscation Grounding (ICLR 2025 · similarity 0.82)
- AlchemistCoder: Harmonizing and Eliciting Code Capability by Hindsight Tuning on Multi-source Data (NeurIPS 2024 · similarity 0.82)
- ChatGPT-Generated Code Assignment Detection Using Perplexity of Large Language Models (Student Abstract) (AAAI 2024 · similarity 0.81)
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.
