CircuitSeer: Mining High-Quality Data by Probing Mathematical Reasoning Circuits in LLMs
"数据选择" 相关文章
CircuitSeer: Mining High-Quality Data by Probing Mathematical Reasoning Circuits in LLMs
Holdout-Loss-Based Data Selection for LLM Finetuning via In-Context Learning
Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?
mR3: Multilingual Rubric-Agnostic Reward Reasoning Models
Train on Validation (ToV): Fast data selection with applications to fine-tuning
Annotation-Efficient Active Test-Time Adaptation with Conformal Prediction
Data Uniformity Improves Training Efficiency and More, with a Convergence Framework Beyond the NTK Regime
Data-Efficient Training by Evolved Sampling
Autoguided Online Data Curation for Diffusion Model Training
不靠海量数据，如何精准喂养大模型？上交Data Whisperer：免训练数据选择法，10%数据逼近全量效果
UPCORE: Utility-Preserving Coreset Selection for Balanced Unlearning
Teaching LLMs According to Their Aptitude: Adaptive Reasoning for Mathematical Problem Solving
TACOS: Open Tagging and Comparative Scoring for Instruction Fine-Tuning Data Selection
MASS: Mathematical Data Selection via Skill Graphs for Pretraining Large Language Models
ACL 2025 | 数据多不如风格齐？SCAR精选<1%样本，指令微调效果飙升
字节最新大模型秘籍：只挑能有推理潜力的数据训练！1.3B模型无需标签自动挑选
Model Performance Begins with Data: Researchers from Ai2 Release DataDecide—A Benchmark Suite to Understand Pretraining Data Impact Across 30K LLM Checkpoints
10篇R1相关的研究全面汇总，万字思考！
Enhancing Instruction Tuning in LLMs: A Diversity-Aware Data Selection Strategy Using Sparse Autoencoders
