Practical Lessons from Reinforcement Learning Post Training Experiments [pdf]

luvverma20111 pts0 comments

What Broke & Why: Practical Lessons from Reinforcement Learning Post-Training | Zenodo

Skip to main

You are using an outdated browser. Please upgrade your browser to improve your experience.

Published July 1, 2026

| Version 1.0

Book

Open

What Broke & Why: Practical Lessons from Reinforcement Learning Post-Training

Authors/Creators

Verma, Luv

Description

A practitioner's field guide to reinforcement learning post-training on small GPU budgets (one to eight card, H100-80GB). Written failure-first: each lesson comes from a training run that broke in a specific way, with every load-bearing number tied to a logged evaluation condition or artifact rather than asserted.

The book is organized in three layers. The Journey is the sequence of programs actually run — math, search, entropy, mixture-of-experts, SWE-bench, and distillation — each of which failed in a way that forced the next. The Science is what those runs proved about how RL changes a model. The Reference is a symptom-indexed field guide for debugging a failing run, including a catalog of failure modes with their fixes.

Topics include verifiable-reward RL, reward design and reward hacking, entropy collapse and training stability (Clip-Cov, GSPO), correctness-gated rewards, MoE routing under RL, evaluation discipline and small-eval noise, the rollout engine, and disaggregated inference. The work is independent and grounded throughout in real training logs and experimental artifacts.

Files

RL_Book.pdf

Files<br>(3.1 MB)

Name<br>Size

Download all

RL_Book.pdf

md5:901bb271a46580b279b4164be177f6a6

3.1 MB

Preview

Download

13

Views

Downloads

Show more details

All versions<br>This version

Views

Total views

13

13

Downloads

Total downloads

Data volume

Total data volume

9.3 MB<br>9.3 MB

More info on how stats are collected....

Versions

External resources

Indexed in

OpenAIRE

Communities

Keywords and subjects

Keywords

RLHF

Reinforcement Learning

post-training

GRPO

GSPO

reward modeling

LoRA

entropy collapse

Details

DOI

DOI Badge

DOI

10.5281/zenodo.21115798

Markdown

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.21115798.svg)](https://doi.org/10.5281/zenodo.21115798)

reStructuredText

.. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.21115798.svg<br>:target: https://doi.org/10.5281/zenodo.21115798

HTML

Image URL

https://zenodo.org/badge/DOI/10.5281/zenodo.21115798.svg

Target URL

https://doi.org/10.5281/zenodo.21115798

Resource type<br>Book

Publisher<br>Zenodo

Languages

English

Rights

License

Creative Commons Attribution 4.0 International

The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited.

Read more

Copyright

© 2026 Luv Verma. Licensed under CC BY 4.0.

Citation

Export

Technical metadata

Created

July 1, 2026

Modified

July 1, 2026

Jump up

This site uses cookies. Find out more on how we use cookies

Accept all cookies<br>Accept only essential cookies

zenodo training https reinforcement learning post

Related Articles