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..
Learning the Plasticity: Plasticity-Driven Learning Framework in Spiking Neural Networks
Authors: Guobin Shen, Dongcheng Zhao, Yiting Dong, Yang Li, Feifei Zhao, Yi Zeng
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We utilize a task known as the copying task [46], as illustrated in Fig. 2A. In this task, SNNs are initially presented with a sequence of stimuli... We evaluated our method on five continuous control environments based on the Brax simulator (ant_dir, swimmer_dir, halfcheetah_vel, hopper_vel, ur5e, fetch). |
| Researcher Affiliation | Academia | 1Beijing Institute of AI Safety and Governance (Beijing-AISI) 2Beijing Key Laboratory of Safe AI and Superalignment 3Brain Cog Lab, CASIA 4Long-term AI EMAIL |
| Pseudocode | Yes | Algorithm 1 Parameter-Exploring Policy Gradients (PEPG) |
| Open Source Code | Yes | Justification: The code will be provided in the supplementary material. |
| Open Datasets | Yes | We use the Brax [47] simulator to design six continuous control environments. In these settings, agents need to navigate at various speeds, directions, and destination points. ... The data sets involved in the experiments are all open data sets. |
| Dataset Splits | Yes | The training task set includes 8 directions, uniformly sampled from [0, 360] degrees. As shown in Fig. 3D, the generalization test task set includes 72 directions, uniformly sampled from [0, 360] degrees. ... The training tasks include 8 speeds, uniformly sampled from [1, 10] m/s. The generalization test tasks include 72 different speeds, uniformly sampled from the same range as the training tasks. |
| Hardware Specification | Yes | All experiments were conducted on a server equipped with 8 NVIDIA A100 GPUs, each with 40 GB of memory. |
| Software Dependencies | No | The implementations were carried out using the JAX framework [57]. No specific version number for JAX or other software dependencies are provided in the text. |
| Experiment Setup | Yes | Unless expressly stated otherwise, the parameter settings and their explanations are shown in Table 4. Table 4: Parameters in PEPG. Parameter Value Description θ 0 Initial policy parameters σ 0.1 Initial adaptive noise scaling parameters αθ 0.15 Learning rate for policy parameters ασ 0.1 Learning rate for adaptive noise scaling mθ, vθ 0, 0 Initial Adam parameters for policy parameters mσ, vσ 0, 0 Initial Adam parameters for adaptive noise scaling β1, β2 0.9, 0.999 Hyperparameters of Adam optimizer ϵ 10 8 Adam parameters M 1500 Number of generations N 128 Number of offspring per generation |