[REVIEW]: Mighty: A Comprehensive Tool for studying Generalization, Meta-RL and AutoRL

[editorialbot]

Submitting author: @amsks (Aditya Mohan)
Repository: https://github.com/automl/Mighty/
Branch with paper.md (empty if default branch): joss_paper
Version: v1.0.0
Editor: @abhishektiwari
Reviewers: @janikmu, @kvr06-ai
Archive: Pending

Status

Status badge code:

HTML: <a href="https://joss.theoj.org/papers/f6175921abef082b7924e76190200379"><img src="https://joss.theoj.org/papers/f6175921abef082b7924e76190200379/status.svg"></a>
Markdown: [![status](https://joss.theoj.org/papers/f6175921abef082b7924e76190200379/status.svg)](https://joss.theoj.org/papers/f6175921abef082b7924e76190200379)

Reviewers and authors:

Please avoid lengthy details of difficulties in the review thread. Instead, please create a new issue in the target repository and link to those issues (especially acceptance-blockers) by leaving comments in the review thread below. (For completists: if the target issue tracker is also on GitHub, linking the review thread in the issue or vice versa will create corresponding breadcrumb trails in the link target.)

Reviewer instructions & questions

@janikmu & @kvr06-ai, your review will be checklist based. Each of you will have a separate checklist that you should update when carrying out your review.
First of all you need to run this command in a separate comment to create the checklist:

@editorialbot generate my checklist

The reviewer guidelines are available here: https://joss.readthedocs.io/en/latest/reviewer_guidelines.html. Any questions/concerns please let @abhishektiwari know.

✨ Please start on your review when you are able; the expectation is that it will take 2-4 weeks to complete. ✨

Checklists

📝 Checklist for @kvr06-ai

📝 Checklist for @janikmu

[editorialbot]

Hello humans, I'm @editorialbot, a robot that can help you with some common editorial tasks.

For a list of things I can do to help you, just type:

@editorialbot commands

For example, to regenerate the paper pdf after making changes in the paper's md or bib files, type:

@editorialbot generate pdf

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

✅ OK DOIs

- 10.52202/079017-2801 is OK
- 10.36227/techrxiv.23739099.v2 is OK
- 10.1109/iros45743.2020.9341462 is OK
- 10.1613/jair.1.14174 is OK
- 10.1145/3583133.3590733 is OK
- 10.1613/jair.1.15703 is OK
- 10.1613/jair.1.13596 is OK
- 10.1109/iros.2017.8202133 is OK

🟡 SKIP DOIs

- No DOI given, and none found for title: Agent57: Outperforming the atari human benchmark
- No DOI given, and none found for title: A survey of meta-reinforcement learning
- No DOI given, and none found for title: Contextualize me – the case for context in reinfor...
- No DOI given, and none found for title: Torchrl: A data-driven decision-making library for...
- No DOI given, and none found for title: Large-scale study of curiosity-driven learning
- No DOI given, and none found for title: Temporally-extended ε-greedy exploration
- No DOI given, and none found for title: Self-paced context evaluation for contextual reinf...
- No DOI given, and none found for title: Hyperparameters in reinforcement learning and how ...
- No DOI given, and none found for title: Model-agnostic meta-learning for fast adaptation o...
- No DOI given, and none found for title: Soft actor-critic: Off-policy maximum entropy deep...
- No DOI given, and none found for title: Cleanrl: High-quality single-file implementations ...
- No DOI given, and none found for title: Open rl benchmark: Comprehensive tracked experimen...
- No DOI given, and none found for title: Prioritized level replay
- No DOI given, and none found for title: Learning quadrupedal locomotion over challenging t...
- No DOI given, and none found for title: RLlib: Abstractions for distributed reinforcement ...
- No DOI given, and none found for title: SGDR: Stochastic gradient descent with warm restar...
- No DOI given, and none found for title: Human-level control through deep reinforcement lea...
- No DOI given, and none found for title: Autorl hyperparameter landscapes
- No DOI given, and none found for title: Stable-baselines3: Reliable reinforcement learning...
- No DOI given, and none found for title: Prioritized experience replay
- No DOI given, and none found for title: Proximal policy optimization algorithms
- No DOI given, and none found for title: Mastering the game of go with deep neural networks...
- No DOI given, and none found for title: Pufferlib 2.0: Reinforcement learning at 1m steps/...
- No DOI given, and none found for title: Syllabus: Portable curricula for reinforcement lea...
- No DOI given, and none found for title: Stoix: Distributed single-agent reinforcement lear...
- No DOI given, and none found for title: Gymnasium: A standard interface for reinforcement ...
- No DOI given, and none found for title: Deep reinforcement learning with double q-learning
- No DOI given, and none found for title: A super-human vision-based reinforcement learning ...
- No DOI given, and none found for title: Tianshou: A highly modularized deep reinforcement ...
- No DOI given, and none found for title: RLlib flow: Distributed reinforcement learning is ...
- No DOI given, and none found for title: Hydra - a framework for elegantly configuring comp...
- No DOI given, and none found for title: Noveld: A simple yet effective exploration criteri...

❌ MISSING DOIs

- 10.52202/068431-1197 may be a valid DOI for title: Discovered policy optimisation

❌ INVALID DOIs

- None

[editorialbot]

👉📄 Download article proof 📄 View article proof on GitHub 📄 👈

[editorialbot]

Repository Analysis Report

Code breakdown

github.com/AlDanial/cloc v 1.98  T=2.08 s (105.3 files/s, 601082.2 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
CSV                             29              0              0        1227817
Python                          81           2152           1474          11528
YAML                            77            337            132           4032
Markdown                        17            260             29           1171
TeX                              1             40              0            325
SVG                              2             53              0            241
Jupyter Notebook                 2              0            334            113
TOML                             1             10              1             88
make                             1             20             11             68
Bourne Shell                     1              3              3              8
JSON                             7              0              0              7
-------------------------------------------------------------------------------
SUM:                           219           2875           1984        1245398
-------------------------------------------------------------------------------

Commit Count by Author

339	Theresa Eimer
   133	Aditya Mohan
    83	Carolin Benjamins
    23	Andre Biedenkapp
    21	André Biedenkapp
    13	amsks
     8	benjamc
     1	AndreBiedenkapp

Repository History

Repository age: First commit on May 27, 2021

Code distribution (top 3 48-hour windows):

Period	Insertions	% of Total	Signal	View Changes
Nov 17, 2025 - Nov 19, 2025	755413	36.0%	🟡	View diff
Aug 05, 2025 - Aug 07, 2025	754632	35.9%	🟡	View diff
May 28, 2025 - May 30, 2025	478737	22.8%	🟢	View diff

GitHub Activity Metrics

Metric	Count
GitHub stars	58
Forks	1
GitHub contributors	5
Releases	1
Total issues	65
Total pull requests	50
Unique commenters/reviewers (excluding authors)	7

[editorialbot]

Paper file info:

📄 Wordcount for paper.md is 1758

✅ The paper includes a Statement of need section

✅ The paper includes a State of the field section

✅ The paper includes a Software design section

✅ The paper includes a Research impact statement section

✅ The paper includes an AI usage disclosure section

[editorialbot]

License info:

✅ License found: BSD 3-Clause "New" or "Revised" License (Valid open source OSI approved license)

[kvr06-ai]

Review checklist for @kvr06-ai

AI usage

• I confirm that I have read the JOSS AI policy and understand that:
  • AI may only be used for non-substantive assistance (e.g., grammar checking my review text, contextualizing public materials)
  • All evaluative decisions—including scoring, accept/reject recommendations, and assessments of originality, correctness, significance, and ethics compliance—must be made exclusively by me, not AI
  • I bear full responsibility for all judgments, recommendations, and technical claims in my review
  • I will disclose any AI assistance at the conclusion of my review

Conflict of interest

• I confirm that I have read the JOSS conflict of interest (COI) policy and that: I have no COIs with reviewing this work or that any perceived COIs have been waived by JOSS for the purpose of this review.

Code of Conduct

• I confirm that I read and will adhere to the JOSS code of conduct.

General checks

• Repository: Is the source code for this software available at the https://github.com/automl/Mighty/?
• License: Does the repository contain a plain-text LICENSE or COPYING file with the contents of an OSI approved software license?
• Contribution and authorship: Has the submitting author (@amsks) made major contributions to the software? Does the full list of paper authors seem appropriate and complete?
• Scope and significance: Does this submission demonstrate clear research impact or credible scholarly significance as described in the JOSS guidelines?
• Data sharing: If the paper contains original data, data are accessible to the reviewers. If the paper contains no original data, please check this item.
• Reproducibility: If the paper contains original results, results are entirely reproducible by reviewers. If the paper contains no original results, please check this item.
• Human and animal research: If the paper contains original data research on humans subjects or animals, does it comply with JOSS's human participants research policy and/or animal research policy? If the paper contains no such data, please check this item.

Development history and open-source practice

• Development timeline: Does the project show evidence of sustained development over time (preferably months or years) rather than rapid, recent code generation?
• Open development: Was the software developed openly from early stages? For projects with recent public repositories, is there at least six months of public development history with evidence of releases, public issues/pull requests, and ideally external engagement?
• Collaborative effort: Does the project show evidence of community engagement such as contributions from multiple developers, issues/discussions from external users, or community influence on the development described in the paper?
• Good practices: Does the project look like a reusable, non-throwaway research software project that follows good open-source practices (license, documentation, tests and/or verification processes, releases, and clear contribution/support pathways).

Functionality

• Installation: Does installation proceed as outlined in the documentation?
• Functionality: Have the functional claims of the software been confirmed?
• Performance: If there are any performance claims of the software, have they been confirmed? (If there are no claims, please check off this item.)

Documentation

• A statement of need: Do the authors clearly state what problems the software is designed to solve and who the target audience is?
• Installation instructions: Is there a clearly-stated list of dependencies? Ideally these should be handled with an automated package management solution.
• Example usage: Do the authors include examples of how to use the software (ideally to solve real-world analysis problems).
• Functionality documentation: Is the core functionality of the software documented to a satisfactory level (e.g., API method documentation)?
• Automated tests: Are there automated tests or manual steps described so that the functionality of the software can be verified?
• Community guidelines: Are there clear guidelines for third parties wishing to 1. Contribute to the software 2. Report issues or problems with the software 3. Seek support

Software paper

• Summary (required section): Has a clear description of the high-level functionality and purpose of the software for a diverse, non-specialist audience been provided?
• A statement of need (required section): Does the paper have a section titled 'Statement of need' that clearly states what problems the software is designed to solve, who the target audience is, and its relation to other work?
• State of the field (required section): Do the authors describe how this software compares to other commonly-used packages? If related tools exist, do they provide a clear "build vs. contribute" justification explaining their unique scholarly contribution and why alternatives are insufficient?
• Software Design (required section): Does the paper include a "Software Design" section that explains the trade-offs weighed, the design/architecture chosen, and why it matters? Is the content compelling and does it demonstrate meaningful design thinking?
• Research Impact Statement (required section): Does the paper include a "Research Impact Statement" section providing evidence of realized impact (publications, external use, integrations) or credible near-term significance (benchmarks, reproducible materials, community-readiness signals)? Is the evidence presented compelling?
• AI usage disclosure (required section): Does the paper include an "AI usage disclosure" section that transparently discloses any use of generative AI in the software creation, documentation, or paper authoring?
• Quality of writing: Is the paper well written (i.e., it does not require editing for structure, language, or writing quality)?
• References: Is the list of references complete, and is everything cited appropriately that should be cited (e.g., papers, datasets, software)? Do references in the text use the proper citation syntax?

[kvr06-ai]

First-pass review

Setup: installed mighty-rl@v1.0.0 into a clean Python 3.11 uv venv, ran the pytest suite, smoke-tested run_mighty.py (PPO and DQN on CartPole, plus PPO with the RND meta-component), ran mkdocs build --strict, and read the PDF proof.

Recommending Minor Revisions. The four items below are small and easy to address.

Checklist items not yet tickable

1. Installation ("Does installation proceed as outlined in the documentation?")

uv pip install mighty-rl (and make install) resolves gymnasium to 1.2.3. That breaks mighty/mighty_agents/base_agent.py:33, which imports NormalizeObservation and NormalizeReward from gymnasium.wrappers.normalize (removed in gymnasium 1.0). Result: 15 test modules fail at collection and import mighty raises ModuleNotFoundError. Pinning gymnasium<1.0 or updating the import restores a working install. Related: #103, #108.

2. Community guidelines ("Are there clear guidelines for third parties wishing to 1. Contribute… 2. Report issues… 3. Seek support")

The repo has no CONTRIBUTING.md. The PR template at .github/Pull_Request_Template/pull_request_template.md:3 already references the missing file ("please ensure you have taken a look at the contribution guidelines in the CONTRIBUTING.md"). Issue reporting is covered by the issue template, but contribution and support pathways are not.

3. Good practices ("…clear contribution/support pathways")

Same gap as item 2. This item becomes tickable once CONTRIBUTING.md (and ideally CODE_OF_CONDUCT.md) is added.

4. Reproducibility ("original results are entirely reproducible by reviewers")

The CartPole and MountainCar rows in the validation table reproduce from committed configs. mighty/configs/environment/gymnasium/ contains cartpole, mountaincar, mountaincarcontinuous, pendulum, and atari_pong only. The SAC rows on Walker2D and HalfCheetah are therefore not directly reproducible. Committing the configs used for those runs would resolve this.

Observations that do not cleanly map to a checklist item

These came up during review but do not map to a specific box.

• docs/index.md:73-75 calls Mighty "early development" and lists Meta-RL and AutoRL as future work, which contradicts the paper. Does not map to Statement of need (docs): the section itself is clear and complete (the issue is editorial drift in one paragraph, not a missing element).
• 10.52202/068431-1197 (editorialbot's candidate DOI for Lu et al., lu-neurips22a) is worth verifying and adding. Does not map to References: the bibliography is complete with proper citation syntax, and an optional DOI is not itself a checklist violation.
• "Cite Us" BibTeX in README.md and docs/index.md:83-88 disagree on authors (docs adds F. Hutter). Does not map to References or Contribution and authorship: this concerns how users cite the software, not the paper's bibliography or its byline.

Positives

• All required paper sections present (Summary, Statement of need, State of the field, Software Design, Research Impact, AI disclosure); word count 1758.
• State of the field positions Mighty against TorchRL, Tianshou, SB3, CleanRL, RLlib, and STOIX with an explicit build-vs-contribute argument.
• BSD-3 (OSI) license; sustained development since May 2021; v1.0.0 released August 2025; one external PR contributor (@jmtoepperwien).
• mkdocs build --strict runs clean; API docs are auto-generated via mkdocstrings.
• With gymnasium pinned, the framework works end-to-end: PPO, DQN, and PPO + RND each train cleanly on CartPole within seconds.

Issue tracking: filed automl/Mighty#116 for items 2–3 (CONTRIBUTING.md / CODE_OF_CONDUCT). Item 1 (gymnasium) is already tracked upstream in #103 and #108. Item 4 (reproducibility) is small enough to address directly from this thread.