DS-1000

The 'DS-1000' dataset/benchmark contains a collection of domain-specific coding tasks used to evaluate the effectiveness of code generation models in generating specialized solutions for real-world software development scenarios.

Papers using DS-1000 (16)

Knowledge-Enhanced Program Repair for Data Science Code2025 · 2 citesDeep-Bench: Deep Learning Benchmark Dataset for Code Generation2025 · 1 citesDomAgent: Leveraging Knowledge Graphs and Case-Based Reasoning for Domain-Specific Code Generation2026WizardCoder: Empowering Code Large Language Models with Evol-Instruct2023 · 82 citesDS-1000: A Natural and Reliable Benchmark for Data Science Code Generation2022 · 32 citesSelfEvolve: A Code Evolution Framework via Large Language Models2023 · 16 citesGrounding Data Science Code Generation with Input-Output Specifications2024 · 2 citesNaturalCodeBench: Examining Coding Performance Mismatch on HumanEval and Natural User Prompts2024 · 2 citesUncovering Weaknesses in Neural Code Generation2024 · 2 citesAn Empirical Study on Self-correcting Large Language Models for Data Science Code Generation2024 · 1 citesTraining Language Models on Synthetic Edit Sequences Improves Code Synthesis2024 · 1 citesCoCoST: Automatic Complex Code Generation with Online Searching and Correctness Testing2024InfiBench: Evaluating the Question-Answering Capabilities of Code Large Language Models2024XFT: Unlocking the Power of Code Instruction Tuning by Simply Merging Upcycled Mixture-of-Experts2024InverseCoder: Self-improving Instruction-Tuned Code LLMs with Inverse-Instruct2024WizardCoder: Empowering Code Large Language Models with Evol-Instruct2023