# Safe Reinforcement Learning on Camera Dropbox

Source: [Safe Reinforcement Learning on Camera Dropbox](<https://www.sota2.com/research/sota/safe-reinforcement-learning-on-camera-dropbox>)

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

Data updated: 2026-04-01T09:05:10.400Z

Evaluation of Safe Reinforcement Learning performance using the Camera Dropbox environment, measured by reward hacking rate, intended behavior rate, failure rate, and true return.

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 Reinforcement Learning |
| Dataset (including any stated split) | Camera Dropbox |
| Metric | Reward Hacking Rate |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 3 of 3 |
| 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

- [Reward Hacking Rate](<https://www.sota2.com/research/sota/safe-reinforcement-learning-on-camera-dropbox.md?metric=Reward%20Hacking%20Rate>): Lower is better; unit: PERCENT.

- [Intended Behavior Rate](<https://www.sota2.com/research/sota/safe-reinforcement-learning-on-camera-dropbox.md?metric=Intended%20Behavior%20Rate>): Higher is better; unit: PERCENT.

- [Failure Rate](<https://www.sota2.com/research/sota/safe-reinforcement-learning-on-camera-dropbox.md?metric=Failure%20Rate>): Lower is better; unit: PERCENT.

- [True Return](<https://www.sota2.com/research/sota/safe-reinforcement-learning-on-camera-dropbox.md?metric=True%20Return>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Reward Hacking Rate | Result date | Report paper |
| --- | --- | --- | --- |
| Oracle MONA | 0 | Not recorded | 69cca6ac0636e5bf4afcbcf9 |
| Best Learned | 0 | Not recorded | 69cca6ac0636e5bf4afcbcf9 |
| Ordinary RL | 91.5 | Not recorded | 69cca6ac0636e5bf4afcbcf9 |

### Oracle MONA

Reward Hacking Rate: 0. Result ID: 69ccdfb147f34ce924b6c3e5.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6ac0636e5bf4afcbcf9.md>)

Recorded evaluation configuration: {"Source":"public ref."}. This record does not assert independent reproduction.

### Best Learned

Reward Hacking Rate: 0. Result ID: 69ccdfb147f34ce924b6c3e6.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6ac0636e5bf4afcbcf9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cca6ac0636e5bf4afcbcf9.md>)

Recorded evaluation configuration: {"Source":"local pilot","Horizon (h)":"1","Calibration type":"sigmoid","Dataset size":"512","PPO steps budget":"1536"}. This record does not assert independent reproduction.

### Ordinary RL

Reward Hacking Rate: 91.5. Result ID: 69ccdfb147f34ce924b6c3e4.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6ac0636e5bf4afcbcf9.md>)

Recorded evaluation configuration: {"Source":"public ref."}. This record does not assert independent reproduction.

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