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..
On Feasible Rewards in Multi-Agent Inverse Reinforcement Learning
Authors: Till Freihaut, Giorgia Ramponi
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The paper is mainly theoretical. However, simple empirical validations are given and the details for these examples can be found in Appendix G. ... We observe that while BC may perform better in the original environment, for the first iterations, the Uniform Sampling Algorithm proves superior when transferring the reward function, especially as the number of samples increases and the environment changes. |
| Researcher Affiliation | Academia | Till Freihaut Department of Computer Science University of Zurich EMAIL Giorgia Ramponi Department of Computer Science University of Zurich EMAIL |
| Pseudocode | Yes | Algorithm 1 MAIRL Uniform Sampling Algorithm with Generative Model |
| Open Source Code | No | The full code will be released after publication. |
| Open Datasets | No | The paper uses a "3 x 3 Gridworld example" and environments "similar to the ones used in Hu and Wellman [2003]". It does not explicitly state that a publicly available dataset is used or provide access information for it. |
| Dataset Splits | No | The paper describes numerical verifications in simulated environments (e.g., "10000 iterations", "allocating the samples uniformly over S A B") rather than using predefined training/test/validation splits on a fixed dataset. There is no mention of explicit percentages or counts for dataset splits. |
| Hardware Specification | No | No specific requirements are needed regarding compute resources as the experiments are only on a small scale. |
| Software Dependencies | No | The paper mentions "Nash Q-Learning" as a method but does not list any specific software libraries, frameworks, or solvers with version numbers that are critical for reproducibility. |
| Experiment Setup | No | The paper describes the environment setup for numerical verifications (e.g., "simple algorithm that iteratively computes the expected reward," "repeated until the strategies... are not changing anymore," "10000 iterations"). However, it does not provide specific numerical hyperparameters such as learning rates, batch sizes, discount factor values, or optimizer settings for any of the algorithms used in the numerical examples. |