Demystifying Reinforcement Learning Post-Training of Language Models
cs.LG, cs.AI, cs.CL
Submitted: 2026-08-24
Updated: 2026-08-28
Comments: Link to website: https://minjang10.github.io/demystifying-rl-finetuning-web Link to code: https://github.com/sankarh-1/demystifying-rl-finetuning
Code: https://github.com/sankarh-1/demystifying-rl-finetuning
Project page: https://minjang10.github.io/demystifying-rl-finetuning-web
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks
- Deep reinforcement learning from human preferences
- Diversity is All You Need: Learning Skills without a Reward Function
- OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework
- Sequence Tutor: Conservative Fine-Tuning of Sequence Generation Models with KL-control
- Rethinking RL for LLM Reasoning: It's Sparse Policy Selection, Not Capability Learning
- Reinforcement Learning Can Amplify Emergent Misalignment from Harmless Rewards
- Constitutional AI: Harmlessness from AI Feedback
- The Art of Scaling Reinforcement Learning Compute for LLMs
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training
- The Coverage Principle: How Pre-Training Enables Post-Training
- Extreme Parkour with Legged Robots
- Let's Verify Step by Step
- Reward Under Attack: Analyzing the Robustness and Hackability of Process Reward Models
- Training language models to follow instructions with human feedback
- Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards
- Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them
- Learning Dynamics of LLM Finetuning
- Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning
- KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills
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: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks