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..
Multi-Agent Learning under Uncertainty: Recurrence vs. Concentration
Authors: Kyriakos Lotidis, Panayotis Mertikopoulos, Nicholas Bambos, Jose Blanchet
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To further explore the behavior of (FTRL) under different noise levels and step sizes, we conduct an additional set of experiments summarized in Figs. 3 and 4. These figures illustrate the distance from x of the final iterate and the hitting time in a neighborhood of x with varying radii. Specifically, we consider step sizes γ {0.01, 0.02, 0.05, 0.1, 0.2, 0.5} and stochastic feedback of the form v̂t = v(Xt) + σωtfor noise levels σ {0.01, 0.05, 0.1, 0.5, 1}. For each (γ, σ) configuration, we perform 100 independent runs, each consisting of 10,000 iterations. The initial state Y0 in each run is drawn uniformly at random from [0, 1]2. The first plot reports the average final distance of the iterates from the equilibrium, averaged across the 100 runs, while the subsequent plots show the hitting time required for the iterates to enter a neighborhood of the equilibrium of radius r {0.005, 0.01, 0.05, 0.1}. |
| Researcher Affiliation | Academia | Kyriakos Lotidis Stanford University EMAIL Panayotis Mertikopoulos Univ. Grenoble Alpes, CNRS, Inria, Grenoble INP LIG 38000 Grenoble, France EMAIL Nicholas Bambos Stanford University EMAIL Jose Blanchet Stanford University EMAIL |
| Pseudocode | No | The paper describes algorithms (S-FTRL, FTRL) using mathematical equations and descriptions, but it does not present them in a structured pseudocode block or algorithm environment. |
| Open Source Code | Yes | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: The code is included in the supplemental material. |
| Open Datasets | No | The paper does not use publicly available datasets. It defines and simulates specific game models for its experiments: "Strongly monotone games. We consider the strongly monotone two-player min-max game defined by f: [0, 1] [0, 1] R with f(x1, x2) = (x1 - 0.5)^2 + 0.5x1x2 + 2(x2 - 0.5)^2" and "Null-monotone games. Fig. 5 shows the empirical distribution of the final iterates under the (FTRL) dynamics in the classic matching pennies game with entropic regularization, played over the probability simplex with payoff matrix P= (+1, -1) (-1, +1) (-1, +1) (+1, -1)". |
| Dataset Splits | No | The paper describes experiments based on simulated game models rather than traditional datasets, therefore, the concept of training/test/validation dataset splits is not applicable in the context of this work. |
| Hardware Specification | No | The paper does not explicitly mention the specific hardware used for running its experiments. |
| Software Dependencies | No | The paper does not explicitly provide specific software versions for its implementation (e.g., Python 3.x, PyTorch 1.x). |
| Experiment Setup | Yes | To further explore the behavior of (FTRL) under different noise levels and step sizes, we conduct an additional set of experiments summarized in Figs. 3 and 4. These figures illustrate the distance from x of the final iterate and the hitting time in a neighborhood of x with varying radii. Specifically, we consider step sizes γ {0.01, 0.02, 0.05, 0.1, 0.2, 0.5} and stochastic feedback of the form v̂t = v(Xt) + σωtfor noise levels σ {0.01, 0.05, 0.1, 0.5, 1}. For each (γ, σ) configuration, we perform 100 independent runs, each consisting of 10,000 iterations. The initial state Y0 in each run is drawn uniformly at random from [0, 1]2. |