MIKASA-Robo

What the protocol tests

The original reinforcement-learning release and the later VLA expansion have different task collections. The repository keeps a dedicated original branch, while the current VLA version adds language instructions and horizon-based splits.

Reading results

The published VLA protocol fixes evaluation episodes and reports short-, medium- and long-horizon performance. Memory-type breakdowns are useful because success on one memory demand need not transfer to another.

Methodology and reuse limits

Do not label an expanded VLA result as an original-release result. Episode seeds and wrappers are part of the evaluation protocol.

Evaluation protocol

Exact version / scope

Original RL release and MIKASA-Robo-VLA expansion

Primary specReviewed 2026-09-28

Sources 1

CognitiveAISystems/MIKASA-Robo official repository

CognitiveAISystems · Publication date not disclosed · Source accessed 2026-09-28

Official README inspected live. Repository code terms do not automatically cover model weights, data, or third-party assets.

Environment

simulation

Primary specReviewed 2026-09-28

Sources 1

CognitiveAISystems/MIKASA-Robo official repository

CognitiveAISystems · Publication date not disclosed · Source accessed 2026-09-28

Official README inspected live. Repository code terms do not automatically cover model weights, data, or third-party assets.

Metric definition

Per-task and per-horizon-split success rate

Primary specReviewed 2026-09-28

Sources 1

MIKASA-Robo-VLA benchmarking protocol

MIKASA-Robo maintainers · Publication date not disclosed · Source accessed 2026-09-28

Task scope

Not publicly disclosed

Embodiment

Tabletop manipulation in ManiSkill/Gymnasium

Primary specReviewed 2026-09-28

Sources 1

CognitiveAISystems/MIKASA-Robo official repository

CognitiveAISystems · Publication date not disclosed · Source accessed 2026-09-28

Official README inspected live. Repository code terms do not automatically cover model weights, data, or third-party assets.

Access & reuse

Access model

Public research documentation and evaluation resources

Primary specReviewed 2026-09-28

Sources 1

CognitiveAISystems/MIKASA-Robo official repository

CognitiveAISystems · Publication date not disclosed · Source accessed 2026-09-28

Official README inspected live. Repository code terms do not automatically cover model weights, data, or third-party assets.

License

Not publicly disclosed

Additional disclosed fields

Category

Robot evaluation benchmark

Primary specReviewed 2026-09-28

Sources 1

CognitiveAISystems/MIKASA-Robo official repository

CognitiveAISystems · Publication date not disclosed · Source accessed 2026-09-28

Official README inspected live. Repository code terms do not automatically cover model weights, data, or third-party assets.

Reported benchmark results

Relationships & deployments

Related reading is an editorial crosslink. Sourced connections describe relationships reported in the cited material. A link to a versioned profile does not establish compatibility with that version unless the connection note explicitly identifies it.

Record history & verified changes

A research review records when we checked a source. It does not mark a product launch or a new deployment.

Initial reviewed record. No subsequent field change has been recorded.