# Safe Multi-Agent Reinforcement Learning on Pistonball 10 agents

Source: [Safe Multi-Agent Reinforcement Learning on Pistonball 10 agents](<https://www.sota2.com/research/sota/safe-multi-agent-reinforcement-learning-on-pistonball-10-agents>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:39:47.493Z

Evaluation of Safe Multi-Agent Reinforcement Learning performance on the Pistonball 10 agents environment, typically measured by Episodic Return and Constraint Violation rates.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Safe Multi-Agent Reinforcement Learning |
| Dataset (including any stated split) | Pistonball 10 agents |
| Metric | Constraint Violation Rate |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Constraint Violation Rate](<https://www.sota2.com/research/sota/safe-multi-agent-reinforcement-learning-on-pistonball-10-agents.md?metric=Constraint%20Violation%20Rate>): Lower is better; unit: PERCENT.

- [Episodic Return](<https://www.sota2.com/research/sota/safe-multi-agent-reinforcement-learning-on-pistonball-10-agents.md?metric=Episodic%20Return>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Constraint Violation Rate | Result date | Report paper |
| --- | --- | --- | --- |
| Scalable Primal-Dual Actor-Critic | 4.919 | Not recorded | 68fa5a6f1e209131ce9bfdd0 |
| MAPPO-L | 6.884 | Not recorded | 68fa5a6f1e209131ce9bfdd0 |
| Decentralized MAPPO-L | 9.303 | Not recorded | 68fa5a6f1e209131ce9bfdd0 |
| Decentralized Aggregate MAPPO-L | 21.79 | Not recorded | 68fa5a6f1e209131ce9bfdd0 |

### Scalable Primal-Dual Actor-Critic

Constraint Violation Rate: 4.919. Result ID: 698bacaa7a967eddb770814d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5a6f1e209131ce9bfdd0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5a6f1e209131ce9bfdd0.md>)

Recorded evaluation configuration: {"Algorithm":"Ours"}. This record does not assert independent reproduction.

### MAPPO-L

Constraint Violation Rate: 6.884. Result ID: 698bacaa7a967eddb770814e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5a6f1e209131ce9bfdd0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc71e209131cea3042c.md>)

Recorded evaluation configuration: {"Information Access":"Global"}. This record does not assert independent reproduction.

### Decentralized MAPPO-L

Constraint Violation Rate: 9.303. Result ID: 698bacaa7a967eddb7708150.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5a6f1e209131ce9bfdd0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc71e209131cea3042c.md>)

Recorded evaluation configuration: {"Information Access":"Local neighborhood"}. This record does not assert independent reproduction.

### Decentralized Aggregate MAPPO-L

Constraint Violation Rate: 21.79. Result ID: 698bacaa7a967eddb770814f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5a6f1e209131ce9bfdd0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc71e209131cea3042c.md>)

Recorded evaluation configuration: {"Information Access":"Local neighborhood","Reward structure":"Sum of rewards in local neighborhood"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.