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..
CONGO: Compressive Online Gradient Optimization
Authors: Jeremy Carleton, Prathik Vijaykumar, Divyanshu Saxena, Dheeraj Narasimha, Srinivas Shakkottai, Aditya Akella
ICLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical simulations and realworld microservices benchmarks demonstrate CONGO s superiority over gradient descent approaches that do not account for sparsity. |
| Researcher Affiliation | Academia | 1 Texas A&M University, 2 The University of Texas at Austin, 3 Inria |
| Pseudocode | Yes | Algorithm 1 CONGO-E: Compressive Online Gradient Optimization Efficient Version |
| Open Source Code | Yes | Code available at https://github.com/5-Jeremy/CONGO |
| Open Datasets | Yes | We utilize the Social Network application from the Death Star Bench suite (Gan et al., 2019) which represents a small-scale social media platform with various request types such as compose-post , read-user-timeline , and read-home-timeline . |
| Dataset Splits | No | The paper describes simulation scenarios and workload patterns (e.g., fixed workload, variable arrival rate, variable job type) rather than explicit training/test/validation splits for static datasets. While the PPO agent baseline mentions 30 training and 30 testing iterations, this refers to the training/testing of the RL agent itself, not a dataset split for the overall methodology. |
| Hardware Specification | Yes | The numerical simulations were run on a machine with an Intel core i7 processor and an NVIDIA Ge Force RTX 3050 Ti Laptop GPU. CPU: AMD Ryzen Threadripper 3960X 24-Core Processor CPU: Intel(R) Core(TM) i9-9940X CPU @ 3.30GHz 2 x NVIDIA Ge Force RTX 2080 Ti |
| Software Dependencies | No | The paper mentions several software components like 'Numpy', 'pyproximal', 'queueing-tool (Jordon, 2023)', 'scipy optimize Python package', and 'wrk2 tool', but it does not specify version numbers for any of these. |
| Experiment Setup | Yes | Table 1: Hyperparameters for Numerical Experiments Table 6: Glossary of Hyperparameters for Deathstar Bench Trials Table 7: NSGD Hyperparameters for Deathstar Bench Trial Table 8: SGDSP Hyperparameters for Deathstar Bench Trial Table 9: Proximal Policy Optimization (PPO) for Deathstar Bench Trial Table 10: CONGO B Hyperparameters for Deathstar Bench Trial Table 11: CONGO Z Hyperparameters for Deathstar Bench Trial Table 12: CONGO E Hyperparameters for Deathstar Bench Trial |