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..
Cooperative Bargaining Games Without Utilities: Mediated Solutions from Direction Oracles
Authors: Kushagra Gupta, Surya Murthy, Mustafa Karabag, Ufuk Topcu, David Fridovich-Keil
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We prove that unlike popular approaches such as the Nash and Kalai-Smorodinsky bargaining solutions, our approach is invariant to monotonic nonaffine transformations, and that under strong convexity and smoothness assumptions, this approach enjoys global asymptotic convergence to Pareto stationary solutions. Moreover, we show that the bargaining solutions found by our algorithm also satisfy the axioms of symmetry and (under slightly stronger conditions) independence of irrelevant alternatives, which are popular in the literature. Finally, we conduct experiments in two domains, multi-agent formation assignment and mediated stock portfolio allocation, which validate these theoretical results. |
| Researcher Affiliation | Academia | Kushagra Gupta The University of Texas at Austin EMAIL Surya Murthy The University of Texas at Austin EMAIL Mustafa O. Karabag The University of Texas at Austin EMAIL Ufuk Topcu The University of Texas at Austin EMAIL David Fridovich-Keil The University of Texas at Austin EMAIL |
| Pseudocode | No | The paper describes the Di BS algorithm and its iterations mathematically in Section 3, but does not present it as a structured pseudocode or algorithm block. |
| Open Source Code | Yes | Project Website: https://kaugsrha.github.io/dibs-on-neurips. |
| Open Datasets | Yes | Real life stock data was procured sing the yfinance Python package [2] (under the Apache license). We ensured that the simplex constraints for this example were met by using the following strategies: 1. Projecting all agent gradients onto the simplex before performing an update. 2. Shrinking the step size by a factor of 10 if a step would cause any element in the state to become less than zero. If the step size becomes less than 10 12, we stop the algorithm. The initial step size was set to be 0.01. |
| Dataset Splits | No | The paper mentions sampling 100 scenarios for investor profiles but does not specify traditional training/test/validation splits typically used for machine learning datasets. The scenarios are for evaluation rather than splitting a dataset for model training. |
| Hardware Specification | Yes | All experiments were run on a desktop with a 12th Gen Intel(R) Core(TM) i7-12700 12-core CPU. |
| Software Dependencies | No | The paper mentions the 'yfinance Python package [2]' but does not specify its version number, nor does it list version numbers for other software components like Python itself, or other libraries that might be used. |
| Experiment Setup | Yes | Di BS and NBS were both run for 5000 iterations. [...] The initial step size was set to be 0.01. [...] For terminating the algorithms, we used the termination condition of either the step size reaching 10 12, or the algorithm completing 1000 iterations. |