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].
Trustless Audits without Revealing Data or Models
Authors: Suppakit Waiwitlikhit, Ion Stoica, Yi Sun, Tatsunori Hashimoto, Daniel Kang
ICML 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We now evaluate ZKAUDIT-T, including the performance of performing SGD in ZK-SNARKs, the end-to-end accuracy and costs of ZKAUDIT-T, and the effect of our optimizations. [...] We benchmarked SGD and ZKAUDIT-T on image classification and a recommender system on Movie Lens (Harper & Konstan, 2015). |
| Researcher Affiliation | Academia | 1Stanford University 2UC Berkeley 3University of Chicago 4UIUC. |
| Pseudocode | No | The paper describes its methods and procedures in narrative text and equations, but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks or figures. |
| Open Source Code | Yes | Code. We have anonymized our code here: https://anonymous.4open.science/r/zkml72D8/README.md |
| Open Datasets | Yes | We used the following datasets: 1. dermnet (Shanthi et al., 2020): [...] 2. flowers-102 (Nilsback & Zisserman, 2008): [...] 3. cars (Krause et al., 2013): [...] 4. movielens (Harper & Konstan, 2015): [...] We further evaluated ZKAUDIT-T on CIFAR-10 and MNIST. |
| Dataset Splits | No | The paper mentions training and testing sets, but does not provide explicit details on validation dataset splits (e.g., percentages, sample counts, or methods for creating validation sets). |
| Hardware Specification | Yes | Hardware. We use the Amazon Web Services (AWS) g4dn.8xlarge instance type for all experiments. |
| Software Dependencies | No | The paper mentions software like 'halo2' and 'PyTorch' but does not specify their version numbers or any other software dependencies with version details. |
| Experiment Setup | No | The paper mentions varying hyperparameters and using different Mobile Net configurations, but does not provide specific values for hyperparameters such as learning rate, batch size, or number of epochs, nor does it detail the specific configurations used for each experiment in the main text. |