# Demystifying Reinforcement Learning Post-Training of Language Models

arXiv: https://arxiv.org/abs/2608.24949 · HTML: https://mlsanity.com/paper/2608.24949/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning.
- Builds on 16 of its 27 references that are in the map.
- First posted on arXiv 24 Aug 2026.

## Placed in

- LLM reasoning / [Reinforcement learning for reasoning](https://mlsanity.com/lr/reinforcement-learning-for-reasoning/)

## Builds on (most cited first, 16 of 16)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?](https://mlsanity.com/paper/2504.13837/) (2025)
- [Diversity is All You Need: Learning Skills without a Reward Function](https://mlsanity.com/paper/1802.06070/) (2018)
- [Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining](https://mlsanity.com/paper/2504.07912/) (2025)
- [Sequence Tutor: Conservative Fine-Tuning of Sequence Generation Models with KL-control](https://mlsanity.com/paper/1611.02796/) (2016)
- [KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills](https://mlsanity.com/paper/2506.12851/) (2025)
- [The Coverage Principle: How Pre-Training Enables Post-Training](https://mlsanity.com/paper/2510.15020/) (2025)
- [Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them](https://mlsanity.com/paper/2507.10616/) (2025)
- [Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards](https://mlsanity.com/paper/2509.21882/) (2025)
- [Reward Under Attack: Analyzing the Robustness and Hackability of Process Reward Models](https://mlsanity.com/paper/2603.06621/) (2026)
- [Beyond Reasoning: Reinforcement Learning Unlocks Parametric Knowledge in LLMs](https://mlsanity.com/paper/2605.07153/) (2026)
- [Rethinking RL for LLM Reasoning: It's Sparse Policy Selection, Not Capability Learning](https://mlsanity.com/paper/2605.06241/) (2026)
- [Reinforcement Learning Can Amplify Emergent Misalignment from Harmless Rewards](https://mlsanity.com/paper/2605.31328/) (2026)

## Built on by (most cited first, 2 of 2)

- [RLVR is a Kernel, Not a Function: Statistical Inference for pass@k Crossovers](https://mlsanity.com/paper/2609.22547/) (2026)
- [Probe with Participation Trophies: Random-Reward RL as a Probe of LLM Capability](https://mlsanity.com/paper/2610.01066/) (2026)

## Abstract (arXiv)

Reinforcement learning (RL) post-training has emerged as a powerful framework for enhancing the capabilities of large language models (LLMs), enabling impressive reasoning, math, and coding capabilities. Yet for many researchers and practitioners, the principles behind classical RL remain a "black box". In this work, we deconstruct the RL post-training algorithm, investigating each step to clarify what is actually happening beneath the surface. By isolating the mechanics of RL with Verifiable Rewards in a controlled and simplified environment, we examine how RL outcomes are shaped by the base model's prior distribution, the granularity of the reward signal, the diversity of the prompt distribution, and model scale. We use the entropy of the policy's output distribution as a lens to compare the distributions learned through pretraining, SFT, and RL post-training, revealing how each stage shapes model certainty. Our investigation sheds light on how these choices interact to affect post-training success. For example, we show that the effect of so-called 'spurious rewards' depends on the prompt distribution used for post-training. We also provide insight into why the success of RL post-training depends on whether the base model already places sufficient probability mass on the desired behavior, linking it to the classical concept of exploration in RL. Ultimately, we provide this primer as a resource to those in the NLP community wishing to incorporate RL as a tool in their toolbox.