From Weeks to Hours: Fast and Principled SFT Data Generation for Reasoning LLMs
cs.LG
Submitted: 2026-10-31
Updated: 2026-10-31
Project page: https://bigml-cs-ucla.github.io/SFTProbe-page
Terminology
Sources
- Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data
- Benchmarking Large Language Models on Answering and Explaining Challenging Medical Questions
- The Faiss library
- OpenThoughts: Data Recipes for Reasoning Models
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- PubMedQA: A Dataset for Biomedical Research Question Answering
- Let's Verify Step by Step
- When Less is More: Investigating Data Pruning for Pretraining LLMs at Scale
- In Good GRACEs: Principled Teacher Selection for Knowledge Distillation
- MedXpertQA: Benchmarking Expert-Level Medical Reasoning and Understanding
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis