Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Solving Continuous Control with Episodic Memory
Authors: Igor Kuznetsov, Andrey Filchenkov
IJCAI 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our algorithm on Open AI gym domains and show greater sample-efficiency compared with the state-of-the-art model-free off-policy algorithms. We evaluate our model on a set of Open AI gym environments [Dhariwal et al., 2017] (Figure 1) and show that it achieves greater sample efficiency compared with the state-of-the-art off-policy algorithms (TD3, SAC). Figure 4: Evaluation results on Open AI Gym Benchmark. Table 1: Average return over 10 trials of 200000 time steps. corresponds to a standard deviation over 10 trials. |
| Researcher Affiliation | Academia | Igor Kuznetsov , Andrey Filchenkov ITMO University EMAIL, afilchenkov@itmo.ru, |
| Pseudocode | Yes | Algorithm 1 EMAC |
| Open Source Code | Yes | We open sourced our algorithm to achieve reproducibility. All the codes and learning curves can be accessed at: http://github. com/schatty/EMAC. |
| Open Datasets | Yes | We evaluate our algorithm on a set of Open AI gym domains [Dhariwal et al., 2017] |
| Dataset Splits | No | Evaluation is performed every 1000 steps with the reported value as an average from 10 evaluation episodes from different seeds without any exploration. |
| Hardware Specification | Yes | All our experiments are performed on single 1080ti NVIDIA card. |
| Software Dependencies | No | Networks parameters are updated with Adam optimizer [Kingma and Ba, 2014] with a learning rate of 0.001. |
| Experiment Setup | Yes | Networks parameters are updated with Adam optimizer [Kingma and Ba, 2014] with a learning rate of 0.001. All models consists of two hidden layers, size 256, for an actor and a critic and a rectified linear unit (Re LU) as a nonlinearity. For the first 1000 time steps we do not exploit an actor for action selection and choose the actions randomly for the exploration purpose. |