Learning Planning Abstractions from Language

Authors: Weiyu Liu, Geng Chen, Joy Hsu, Jiayuan Mao, Jiajun Wu

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our models in two different domains: Baby AI (Chevalier-Boisvert et al., 2019), a 2D grid world environment, and Kitchen-Worlds (Yang et al., 2023), a 3D robotic manipulation benchmark. We present the results on generalization experiments in Table 1.
Researcher Affiliation Academia Weiyu Liu1 , Geng Chen1 , Joy Hsu1, Jiayuan Mao2 , Jiajun Wu1 1 Stanford 2 MIT
Pseudocode No The paper describes algorithms in paragraph text but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Project page: https://parl2024.github.io/.
Open Datasets Yes We evaluate our models in two different domains: Baby AI (Chevalier-Boisvert et al., 2019), a 2D grid world environment, and Kitchen-Worlds (Yang et al., 2023), a 3D robotic manipulation benchmark.
Dataset Splits No The paper describes the collection of 'training data' and the initialization of 'testing scenes' for evaluation, but it does not provide explicit percentages or counts for training, validation, and test splits, nor does it explicitly mention a separate validation set split from the main dataset.
Hardware Specification No The paper mentions that the algorithm can be 'efficiently implemented on a GPU' but provides no specific details about the GPU model (e.g., NVIDIA A100), CPU, memory, or other hardware specifications used for experiments.
Software Dependencies No The paper mentions software components such as 'PyBullet physics simulator', 'Point Cloud Transformer (PCT)', 'FiLM layers', and 'GPT-4 LLM', but it does not provide specific version numbers for these software dependencies, which is necessary for reproducible setup.
Experiment Setup Yes Parameters. We provide network and training parameters in Table 3. Table 3: Model Parameters (partial content: Epochs 2000, Optimizer Adam, Learning rate 1e-4, Batch size 32, Learning rate warmup Linear (Tend = 20), Learning rate decay Cosine annealing (Tmax = 2000)).