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..
Boosting the Certified Robustness of L-infinity Distance Nets
Authors: Bohang Zhang, Du Jiang, Di He, Liwei Wang
ICLR 2022 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments show that using the proposed training strategy, the certified accuracy of ℓ -distance net can be dramatically improved from 33.30% to 40.06% on CIFAR-10 (ϵ = 8/255), meanwhile outperforming other approaches in this area by a large margin. |
| Researcher Affiliation | Collaboration | 1Key Laboratory of Machine Perception, MOE, School of Artificial Intelligence, Peking University 2Microsoft Research 3International Center for Machine Learning Research, Peking University |
| Pseudocode | No | The paper describes its methods using mathematical equations and prose but does not include any explicit pseudocode or algorithm blocks. |
| Open Source Code | Yes | Codes are available at https://github.com/zbh2047/L inf-dist-net-v2. |
| Open Datasets | Yes | We evaluate the proposed training strategy on benchmark datasets MNIST and CIFAR-10 to show the effectiveness of ℓ -distance net. |
| Dataset Splits | No | The paper mentions 'training dataset' and 'test dataset' (e.g., Figure 1(b)), and uses standard benchmark datasets (MNIST, CIFAR-10), which often come with predefined splits. However, it does not explicitly state the train/validation/test split percentages, sample counts, or the methodology used to create these splits. |
| Hardware Specification | Yes | All experiments are run for 8 times on a single NVIDIA Tesla-V100 GPU |
| Software Dependencies | Yes | Our experiments are implemented using the Pytorch framework. ... The CUDA version is 11.2. |
| Experiment Setup | Yes | In all experiments, we choose the Adam optimizer with a batch size of 512. The learning rate is set to 0.03 initially and decayed using a simple cosine annealing throughout the whole training process. ... The ℓp-relaxation starts at p = 8 and ends at p = 1000 with p increasing exponentially. Accordingly, the mixing coefficient λ decays exponentially during the ℓp-relaxation process from λ0 to a vanishing value λend. ... All hyper-parameters are provided in Table 3. |