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 [1].
Enabling Optimal Decisions in Rehearsal Learning under CARE Condition
Authors: Wen-Bo Du, Hao-Yi Lei, Lue Tao, Tian-Zuo Wang, Zhi-Hua Zhou
ICML 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments validate the effectiveness and efficiency of our method. We evaluate our proposed approach on two datasets including a synthetic dataset and a real-world dataset. The comparison results are summarized in Tab. 2 and Tab. 3, where the number of observational samples is set to 100. |
| Researcher Affiliation | Academia | 1National Key Laboratory for Novel Software Technology, Nanjing University, China 2School of Artificial Intelligence, Nanjing University, China. Correspondence to: Zhi-Hua Zhou <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Projection Newton for selecting optimal action... Algorithm 2 Closed-form solution for cases where |Y| = 1 |
| Open Source Code | No | The paper mentions using the stable-baselines3 library (Raffin et al., 2021) for RL results, which is a third-party library, not code provided by the authors for their methodology. There is no explicit statement about the release of the authors' own source code or a link to a repository. |
| Open Datasets | Yes | The Bermuda dataset, which records environmental variables in the Bermuda area, is described in ecology research (Courtney et al., 2017), with available generation order of variables (Andersson & Bates, 2018). The parameters of structural equations are derived from fitting linear models on normalized data (Qin et al., 2023). |
| Dataset Splits | No | The paper states 'We repeat the experiment under 100 random seeds for each dataset' and 'the number of observational samples is set to 100', but does not provide specific details on how these samples are split into training, validation, or test sets. |
| Hardware Specification | Yes | The experiments are run on a Nvidia Tesla A100 GPU and two Intel Xeon Platinum 8358 CPUs. |
| Software Dependencies | No | The paper mentions 'the RL results (Fig. 4) are obtained by using the stable-baselines3 library (Raffin et al., 2021)'. However, it does not specify the version number of the stable-baselines3 library or other key software components used in their own implementation. |
| Experiment Setup | Yes | We repeat the experiment under 100 random seeds for each dataset, incluing 3 measures as follows: ... The number of observational samples is set to 100. ... feasible alteration values are set to [−1, 1] for each of them. ... Finally, the hyperparameter τ for previous rehearsal-learning methods is selected as the value that achieves the highest average AUF probability among various candidates. |