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.
Chasing Convex Functions with Long-term Constraints
Authors: Adam Lechowicz, Nicolas Christianson, Bo Sun, Noman Bashir, Mohammad Hajiesmaili, Adam Wierman, Prashant Shenoy
ICML 2024 | Venue PDF | LLM Run Details | Input Tokens: 39,497 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 6,059 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 | In this section, we conduct numerical experiments on synthetic CFL instances. We evaluate ALG1 and CLIP against the offline optimal solution, three heuristics adapted from related work, and the learning-augmented Baseline. |
| Researcher Affiliation | Academia | 1Manning College of Information and Computer Sciences, University of Massachusetts Amherst, USA. 2Computing & Mathematical Sciences, California Institute of Technology, USA. 3Cheriton School of Computer Science, University of Waterloo, Ontario, Canada. 4Computer Science & Artificial Intelligence Laboratory, Massachusetts Institute of Technology, USA.. |
| Pseudocode | Yes | Algorithm 1 Pseudo-cost minimization algorithm (ALG1) ... Algorithm 2 Consistency Limited Pseudo-cost minimization (CLIP) |
| Open Source Code | No | The paper does not provide any explicit statements about making its source code available or links to a code repository for the methodology described. |
| Open Datasets | No | We construct a d-dimensional decision space, where d is picked from the set {5, 7, ... , 21}. ... For a given setting of d, U/L, and β, we generate 1,000 random instances as follows. |
| Dataset Splits | No | The paper mentions generating '1,000 random instances' for experiments but does not specify any training, validation, or test splits for these instances in the traditional sense. The experiments evaluate competitive ratios against an offline optimal solution. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the experiments. |
| Software Dependencies | No | We use CVXPY (Diamond & Boyd, 2016) to compute the offline optimal solution for each instance using a convex optimization solver with access to all cost functions in advance. |
| Experiment Setup | Yes | We construct a d-dimensional decision space, where d is picked from the set {5, 7, ... , 21}. ... We set different cost fluctuation ratios U/L {50, 150, ... , 1250} by setting L and U accordingly, and β is picked from the set β {0, 5, ... , U/2.5}. For each experiment, c(x) = x 1. ... The time horizon T is generated randomly from a uniform distribution on [6, 24]. ... To generate ft, we first draw µt from the uniform distribution on [L, U], and then draw each term of ft from a normal distribution centered at µt with standard deviation σ (i.e., f i t N(µt, σ)). ... In the setting with advice, we obtain simulated advice as follows: Let ξ [0, 1] denote an adversarial factor. |