One-shot Imitation in a Non-Stationary Environment via Multi-Modal Skill

Authors: Sangwoo Shin, Daehee Lee, Minjong Yoo, Woo Kyung Kim, Honguk Woo

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 4. Evaluations In this section, we evaluate the performance of our On IS framework under various configurations of non-stationary environments.
Researcher Affiliation Academia 1Department of Computer Science and Engineering, Sungkyunkwan University, Suwon, Republic of Korea.
Pseudocode Yes Algorithm 1: Learning to transfer skills
Open Source Code No We create an expert dataset for long-horizon, multi-stage Meta-world tasks with diverse robotic manipulation skills, and make it publicly available for other research works. This statement refers to the dataset, not the source code for the On IS framework. No other explicit statement or link for the code is found in the paper.
Open Datasets Yes We create an expert dataset for long-horizon, multi-stage Meta-world tasks with diverse robotic manipulation skills, and make it publicly available for other research works.
Dataset Splits Yes We use 16 tasks for training and use 8 tasks for evaluation, where for each task, 15 episodes are generated. Accordingly, we build a dataset of 15 episodes for each of the 16 tasks in 13 different environment settings with distinct dynamics parameters, containing 3,120 trajectories (by 16 × 15 × 13).
Hardware Specification No The paper states 'We implement our On IS framework using the open-source project Jax (Bradbury et al., 2018).' but does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for the experiments.
Software Dependencies No We implement our On IS framework using the open-source project Jax (Bradbury et al., 2018). This only mentions the framework without a specific version, and no other software dependencies with version numbers are listed.
Experiment Setup Yes Table 6. Hyperparameters for S-On IS