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..
Dual-Space Semantic Synergy Distillation for Continual Learning of Unlabeled Streams
Authors: Donghao Sun, Xi Wang, Xu Yang, Kun Wei, Cheng Deng
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experiments conducted on a variety of benchmarks demonstrate that our proposed method attains state-of-the-art performance, and ablation studies further substantiate the effectiveness and superiority of the proposed method. |
| Researcher Affiliation | Academia | Donghao Sun Xi Wang Xu Yang Kun Wei Cheng Deng School of Electronic Engineering, Xidian University, Xi an 710071, China EMAIL |
| Pseudocode | No | The paper describes the methodology verbally and provides diagrams (e.g., Figure 3), but it does not include a formal pseudocode block or algorithm. |
| Open Source Code | No | We use open datasets (CIFAR100/Image Net-R/CUB200) and models (CLIP/GPT-4) but do not provide code/preprocessed data, though future open-sourcing is planned for reproducibility. |
| Open Datasets | Yes | We conducted experiments on three datasets, including the widely used image classification dataset CIFAR100, the style-varied classification dataset Image Net-R, and the fine-grained dataset CUB200. |
| Dataset Splits | Yes | We partitioned each of the three datasets into 5, 10, and 20 consecutive task streams, with an equal distribution across the tasks. |
| Hardware Specification | Yes | All baselines were also re-run using the same Vi T-L/14 backbone for fair comparison. ... Python 3.8, Pytorch 2.0.1, and a single GPU A6000. |
| Software Dependencies | Yes | Python 3.8, Pytorch 2.0.1, and a single GPU A6000. |
| Experiment Setup | Yes | For both datasets, our pretrained model is the Vi T-L/14 version of CLIP, and we train the model with the Adam optimizer for 30 epochs, with a learning rate of 1 10 3. We use Cosine Schedule to adjust the learning rate. ... In our experiments m = 3 and λ1 = 1 for all datasets and λ2 = 0.03 for Image Net-R and CIFAR100, λ2 = 0.15 for CUB200. |