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..
Variance-Reduced Long-Term Rehearsal Learning with Quadratic Programming Reformulation
Authors: Wen-Bo Du, Tian Qin, Tian-Zuo Wang, Zhi-Hua Zhou
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments validate the effectiveness of our approach. We visualize our approaches using a toy example modeling a simplified Texas Hold em game, followed by evaluations on synthetic and real-world datasets. Our methods are compared against baseline approaches and established rehearsal learning methods, including QWZ23 [6] and MICNS [7]. |
| Researcher Affiliation | Academia | Wen-Bo Du, Tian Qin, Tian-Zuo Wang, Zhi-Hua Zhou National Key Laboratory for Novel Software Technology, Nanjing University, China School of Artificial Intelligence, Nanjing University, China EMAIL |
| Pseudocode | Yes | Algorithm 1 GMu R (Greedy Multi-round Rehearsal) Input: start/end time t0/te, SRM para. Θ, desired region S ... Algorithm 2 Far Mu R (Far-sighted Multi-round Rehearsal) Input: start/end time t0/te, SRM para. Θ, desired region S |
| Open Source Code | Yes | The code and data for the main experimental results are provided in the supplemental material. |
| Open Datasets | Yes | The Bermuda dataset records environmental variables in the Bermuda area and has been widely used in prior research [6, 7, 41, 42]. The generation order of variables in this dataset is recorded [43]... [43] Andreas Andersson and Nicholas Bates. In situ measurements used for coral and reef-scale calcification structural equation modeling including environmental and chemical measurements, and coral calcification rates in bermuda from 2010 to 2012 (BEACON project), 2018. http://lod.bco-dmo.org/id/dataset/720788. |
| Dataset Splits | No | The paper uses observational data for parameter estimation and evaluates performance using Monte Carlo samples, but does not explicitly specify training/test/validation dataset splits for model development or evaluation in the main text or appendices. |
| Hardware Specification | Yes | First, all experiments were run on a Nvidia Tesla A100 GPU and two Intel Xeon Platinum 8358 CPUs. |
| Software Dependencies | No | The paper describes algorithms and estimation methods but does not provide specific software names with version numbers for reproducibility. |
| Experiment Setup | Yes | We visualize our approaches using a toy example modeling a simplified Texas Hold em game, followed by evaluations on synthetic and real-world datasets. Our methods are compared against baseline approaches and established rehearsal learning methods, including QWZ23 [6] and MICNS [7]. ... Experimental details are provided in Appx. D. ... Each value is estimated using 1000 Monte Carlo samples, averaged over 5 random seeds. ... The full results on the two datasets are presented in Tab. 1... Appx. D provides true parameters of the synthetic dataset, with variables in the dataset illustrated in Fig. 9. ... The parameters associated with instantaneous influence relations ((A)) and the noise covariance matrix (Cov[εt]) are estimated by fitting least-squares linear models to the real-world data [41, 43], while the parameters associated with the lagged influence relations (B) are manually determined. |