Notice: This website is continuously updated as new conference proceedings and journal volumes are published. Consequently, the figures presented here may differ from those reported in our publications.
A brief overview of our methodology is provided on the Methods page, along with references to publications documenting our approach and links to the code and data used in each study.
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.
High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach
Authors: Tim Pearce, Alexandra Brintrup, Mohamed Zaki, Andy Neely
ICML 2018 | Venue PDF | LLM Run Details | Input Tokens: 17,880 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 5,536 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Benchmark experiments show the method outperforms current state-of-the-art uncertainty quantification methods, reducing average PI width by over 10%. |
| Researcher Affiliation | Academia | 1Department of Engineering, University of Cambridge, UK 2Alan Turing Institute, UK. Correspondence to: Tim Pearce <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Construction of loss function using basic operations |
| Open Source Code | Yes | Code is made available online6. https://github.com/Tea Pearce |
| Open Datasets | Yes | Experiments were run across ten open-access datasets. |
| Dataset Splits | No | The paper states that experiments were run 'across ten open-access datasets' and mentions a 'validation' phase in relation to the neural network training, but it does not specify explicit training/validation/test dataset splits (e.g., percentages or counts) needed for reproduction. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments, such as CPU or GPU models. |
| Software Dependencies | No | The paper mentions 'modern NN APIs' and hints at Python being used via the GitHub link, but it does not provide specific software names with version numbers (e.g., 'PyTorch 1.9' or 'TensorFlow 2.x'). |
| Experiment Setup | Yes | Setting s = 160 worked well in experiments in section 6, requiring no alteration across datasets. Models were asked to output 95% PIs and used five NNs per ensemble. |