数据集调试论文趋势 2023–2026

主要结论

• 数据集调试论文从 2023 年的 30 篇降到 2026 年的 22 篇，热度在回落。
• NeurIPS 是发数据集调试论文最多的会议，2023–2026 年共 40 篇，占四个会议总量的 34%。
• 交叉方向里涨得最快的是扩散模型：同时属于两个主题的论文从 2023 年的 0 篇增至 2026 年的 5 篇。
• 在「以数据为中心 AI」的 5 个主题里，它的 2026 年论文数排第 4；同组增长最快的合成数据生成是 1.3 倍，数据集调试是 0.7 倍。

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

• NeurIPS 2026Can Folding Models Tell Binders from Bluffers? Evidence from POISK: The Patent-Derived Antibody Dataset
• ICML 2026Exploring 3D Dataset Pruning
• ICLR 2026Slicing Wasserstein over Wasserstein via Functional Optimal Transport
• CVPR 2026Rethinking Dataset Distillation Hard Truths about Soft Labels
• NeurIPS 2025Enhancing Sample Selection Against Label Noise by Cutting Mislabeled Easy Examples
• ICML 2025Lightspeed Geometric Dataset Distance via Sliced Optimal Transport
• ICLR 2025Simulating Training Dynamics to Reconstruct Training Data from Deep Neural Networks
• CVPR 2025Curriculum Coarse-to-Fine Selection for High-IPC Dataset Distillation
• NeurIPS 2024Indoor Air Quality Dataset with Activities of Daily Living in Low to Middle-income Communities
• ICML 2024SPADE: Sparsity-Guided Debugging for Deep Neural Networks
• ICLR 2024Teaching Large Language Models to Self-Debug
• CVPR 2024LUWA Dataset: Learning Lithic Use-Wear Analysis on Microscopic Images
• NeurIPS 2023ECG-QA: A Comprehensive Question Answering Dataset Combined With Electrocardiogram
• ICML 2023Dataset Distillation with Convexified Implicit Gradients
• ICLR 2023Concept-level Debugging of Part-Prototype Networks
• CVPR 2023FLAG3D: A 3D Fitness Activity Dataset With Language Instruction
• NeurIPS 2026CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging
• ICML 2026Magnitude Distance: A Geometric Measure of Dataset Similarity
• ICLR 2026BANZ-FS: BANZSL Fingerspelling Dataset
• CVPR 2026Beyond Soft Label Dataset Distillation via Orthogonal Gradient Matching

全部数据集调试论文

共 24 篇，第 1 / 1 页

1. ICLR 2026BANZ-FS: BANZSL Fingerspelling Dataset
2. ICLR 2026Diffusion Models as Dataset Distillation Priors
3. ICLR 2026Slicing Wasserstein over Wasserstein via Functional Optimal Transport
4. ICLR 2026Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
5. ICLR 2025Distilling Dataset into Neural Field
6. ICLR 2025Equivariant Neural Functional Networks for Transformers
7. ICLR 2025MANTRA: The Manifold Triangulations Assemblage
8. ICLR 2025Simulating Training Dynamics to Reconstruct Training Data from Deep Neural Networks
9. ICLR 2025SWEb: A Large Web Dataset for the Scandinavian Languages
10. ICLR 2025Ward: Provable RAG Dataset Inference via LLM Watermarks
11. ICLR 2024Data Debugging with Shapley Importance over Machine Learning Pipelines
12. ICLR 2024Data Filtering Networks
13. ICLR 2024Embarrassingly Simple Dataset Distillation
14. ICLR 2024Multisize Dataset Condensation
15. ICLR 2024Navigating Dataset Documentations in AI: A Large-Scale Analysis of Dataset Cards on HuggingFace
16. ICLR 2024Self-Supervised Dataset Distillation for Transfer Learning
17. ICLR 2024Teaching Large Language Models to Self-Debug
18. ICLR 2024Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching
19. ICLR 2024Training Unbiased Diffusion Models From Biased Dataset
20. ICLR 2023Concept-level Debugging of Part-Prototype Networks
21. ICLR 2023Dataset Pruning: Reducing Training Data by Examining Generalization Influence
22. ICLR 2023Metadata Archaeology: Unearthing Data Subsets by Leveraging Training Dynamics
23. ICLR 2023Mitigating Dataset Bias by Using Per-Sample Gradient
24. ICLR 2023Transferable Unlearnable Examples

常见问题

2023–2026 年四大会议一共有多少篇数据集调试论文？

共 119 篇，按年份分别是 2023 年 30 篇、2024 年 40 篇、2025 年 27 篇、2026 年 22 篇。

NeurIPS 2026 有多少篇数据集调试论文？

NeurIPS 2026 有 9 篇数据集调试论文，上一届是 9 篇。

ICLR 2026 有多少篇数据集调试论文？

ICLR 2026 有 4 篇数据集调试论文，上一届是 6 篇。

ICML 2026 有多少篇数据集调试论文？

ICML 2026 有 3 篇数据集调试论文，上一届是 2 篇。

CVPR 2026 有多少篇数据集调试论文？

CVPR 2026 有 6 篇数据集调试论文，上一届是 10 篇。

哪个会议的数据集调试论文最多？

NeurIPS，2023–2026 年共 40 篇。