Understanding Reasoning from Pretraining to Post-Training
cs.LG, cs.AI, cs.CL
Submitted: 2026-07-17
Updated: 2026-11-04
Code: https://github.com/pavelslab-nyu/pre2post-chess
Terminology
Sources
- Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data
- RL Excursions during Pre-Training: Re-examining Policy Optimization for LLM training
- Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision
- The Coverage Principle: How Pre-Training Enables Post-Training
- IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL
- Training Verifiers to Solve Math Word Problems
- Training Compute-Optimal Large Language Models
- Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment
- Scaling Laws for Neural Language Models
- Reasoning with Sampling: Your Base Model is Smarter Than You Think
- The Art of Scaling Reinforcement Learning Compute for LLMs
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training
- Large Language Model Guided Tree-of-Thought
- Nemotron-CC-Math: A 133 Billion-Token-Scale High Quality Math Pretraining Dataset
- 2 OLMo 2 Furious
- Olmo 3
- EvoLM: In Search of Lost Language Model Training Dynamics
- Scaling Laws of Synthetic Data for Language Models
- Test-Time Scaling Makes Overtraining Compute-Optimal
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
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