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].
Bridging Fairness and Efficiency in Conformal Inference: A Surrogate-Assisted Group-Clustered Approach
Authors: Chenyin Gao, Peter B. Gilbert, Larry Han
ICML 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, through simulations and an analysis of the phase 3 Moderna COVE COVID-19 vaccine trial, we demonstrate that SAGCCI outperforms existing methods, producing narrower prediction sets while maintaining valid group-conditional coverage, effectively balancing fairness and efficiency in uncertainty quantification. [...] 5. Simulation studies [...] 6. Real-data application |
| Researcher Affiliation | Academia | 1Harvard University, Department of Biostatistics, Boston, MA, USA 2Vaccine and Infectious Disease and Public Health Sciences Divisions, Fred Hutch Cancer Center, Seattle, WA, USA 3Northeastern University, Department of Public Health and Health Sciences, Boston, MA, USA. |
| Pseudocode | Yes | Algorithm 1 Surrogate-Assisted Group Clustered Conformal Inference (SAGCCI) |
| Open Source Code | Yes | Our code is publicly available at https:// github.com/Gaochenyin/Surr Conformal DR. |
| Open Datasets | No | We analyzed data from the Moderna COVE phase 3 COVID-19 vaccine efficacy trial, which randomized adults to receive two doses of m RNA-1273 or placebo at Days 1 and 29. |
| Dataset Splits | Yes | Following Sesia & Cand es (2020), we split 75% of the data into the first fold I1 for model training, and use the remaining 25% as the second fold I2 to construct prediction sets. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The non-conformity scores for the categorical outcomes are derived using the nested prediction sets approach from Kuchibhotla & Berk (2023); more details are presented in the Appendix. ... The models are all fitted using SUPERLEARNER (Van der Laan et al., 2007), with Random Forests and generalized linear models as base learners. |
| Experiment Setup | No | We set the miscoverage level α = 0.05 and employed a 75-25 train-test split, repeating the procedure 100 times to summarize results. |