Research map / Large language models / Language, safety and interpretability
Can Machines Read Coding Manuals Yet? – A Benchmark for Building Better Language Models for Code Understanding
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Fine-tune BERTOverflow on Forum Answer Ranking | BERTOverflow · Forum Answer Ranking (R) | MRR: 0.5910 -> 0.6743 | Helped |
| Fine-tune BERTOverflow on Forum Link Prediction | BERTOverflow · Forum Link Prediction (L) | T: 59.52 -> 180.42 | Helped |
| Fine-tune CodeBERT on Forum to Class Prediction | CodeBERT · Forum to Class Prediction (F) | T: 12.23 -> 50.07 | Helped |
| Fine-tune BERTOverflow on Class Hierarchy Prediction | BERTOverflow · Class Hierarchy Distance Prediction (H) | Pearson r: 0.17 -> 0.34 | Helped |
| Fine-tune BERTOverflow on Class Usage Prediction | BERTOverflow · Class Usage Prediction (U) | Pearson r: 0.37 to 0.52 | Helped |
| Add hierarchy training to usage training (ablation) | BERTOverflow fine-tuned on U · Class Usage Prediction (U) | Pearson r: 0.52 -> 0.61 | Helped |
| Add usage training to hierarchy training (ablation) | BERTOverflow fine-tuned on H · Class Hierarchy Distance Prediction (H) | Pearson r: 0.34 -> 0.29 | Hurt |
| Fine-tune CodeBERT on U (ablation) | CodeBERT · Class Usage Prediction (U) | Pearson r: 0.33 -> 0.30 | Hurt |
| Use BERTOverflow instead of CodeBERT for single-task R fine-tuning (ablation) | Fine-tuned models · Forum Answer Ranking (R) | MRR: 0.6671 -> 0.6743 | Helped |
| Add H and U to R, F, and L multi-task training (ablation) | BERTOverflow · Forum Link Prediction (L) | T: 197.63 -> 198.10 | No clear effect |
| Use multi-task rather than single-task fine-tuning (ablation) | BERTOverflow on Forum Answer Ranking · Forum Answer Ranking (R) | MRR: 0.6743 -> 0.6879 | Helped |
| Use multi-task rather than single-task fine-tuning (ablation) | BERTOverflow on Forum Link Prediction · Forum Link Prediction (L) | T: 180.42 -> 198.10 | Helped |
Closest papers by meaning
- Toward Interactive Understanding of Code APIs (NeurIPS 2026 · similarity 0.88)
- CodeMMLU: A Multi-Task Benchmark for Assessing Code Understanding & Reasoning Capabilities of CodeLLMs (ICLR 2025 · similarity 0.88)
- CodeSteer: Symbolic-Augmented Language Models via Code/Text Guidance (ICML 2025 · similarity 0.87)
- DyCodeEval: Dynamic Benchmarking of Reasoning Capabilities in Code Large Language Models Under Data Contamination (ICML 2025 · similarity 0.86)
- AlchemistCoder: Harmonizing and Eliciting Code Capability by Hindsight Tuning on Multi-source Data (NeurIPS 2024 · similarity 0.86)
- metabench - A Sparse Benchmark of Reasoning and Knowledge in Large Language Models (ICLR 2025 · 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.