Beyond Compression: Diagnosing How Post-Training Changes Mathematical Reasoning
cs.LG, cs.AI
Submitted: 2026-09-29
Updated: 2026-09-29
Sources
- Evaluating Large Language Models Trained on Code
- Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- OpenThoughts: Data Recipes for Reasoning Models
- GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
- s1: Simple test-time scaling
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- HybridFlow: A Flexible and Efficient RLHF Framework
- Language Models are Multilingual Chain-of-Thought Reasoners
- Functional Benchmarks for Robust Evaluation of Reasoning Performance, and the Reasoning Gap
- Qwen3 Technical Report
- LIMO: Less is More for Reasoning
- From $f(x)$ and $g(x)$ to $f(g(x))$: LLMs Learn New Skills in RL by Composing Old Ones
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks