MIKASA-Robo is a benchmark suite containing ready-made datasets and pre-trained oracle agent checkpoints for memory-intensive robotic manipulation tasks. The datasets and checkpoints were created by the author 'avanturist' and were last updated on May 11, 2025. Detailed instructions for dataset collection and descriptions are available on the project's GitHub repository.
Use Cases
- Benchmarking robotic manipulation agents based on the memory-intensive tasks described.
- Training reinforcement learning agents using the provided pre-trained oracle agent checkpoints.
- Collecting new robotic task datasets by following the provided instructions that utilize the oracle checkpoints.
Strengths
- Includes pre-trained oracle agent checkpoints, which can accelerate research and development.
- Provides a dedicated benchmark suite focused on memory-intensive robotic manipulation, a specific and relevant challenge area.
- Last updated on May 11, 2025, indicating recent maintenance.
Limitations
- Description metadata is limited; actual data quality, size, and formats require manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
Provenance
- Source
- GitHub repository by CognitiveAISystems/MIKASA-Robo.
- Collection Method
- Likely generated via simulation or robotic interaction using oracle agents.
- Time Range
- null
- Freshness
- Last updated 2025-05-11 13:21:43; freshness should be verified.
- Geography
- null