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..
Geometry Meets Incentives: Sample-Efficient Incentivized Exploration with Linear Contexts
Authors: Ben Schiffer, Mark Sellke
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4 Experimental Results Finally, we conclude with a simple experiment validating the practicality of the proposed algorithm. We implemented Algorithm 4 and tested on synthetic data (Figure 2). Our experiments focus on the setting where the prior distribution is a d-dimensional Gaussian that is independent across all dimensions and has mean 0.1 in the first dimension and mean 0 in all other dimensions. |
| Researcher Affiliation | Academia | Benjamin Schiffer Department of Statistics Harvard University 1 Oxford St EMAIL Mark Sellke Department of Statistics Harvard University 1 Oxford St EMAIL |
| Pseudocode | Yes | Algorithm 1 BIC Exploration Pseudocode; Algorithm 2 Initial Exploration Pseudocode; Algorithm 3 Exponential Growth Pseudocode |
| Open Source Code | No | 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: [No] Justification: The algorithm is simple to implement and all details are included in the paper. |
| Open Datasets | No | We implemented Algorithm 4 and tested on synthetic data (Figure 2). Our experiments focus on the setting where the prior distribution is a d-dimensional Gaussian that is independent across all dimensions and has mean 0.1 in the first dimension and mean 0 in all other dimensions. |
| Dataset Splits | No | The paper uses synthetic data generated according to a d-dimensional Gaussian distribution. It describes the goal as finding "the number of samples necessary to achieve λ-spectral exploration" rather than using traditional training/test/validation splits on a dataset. |
| Hardware Specification | Yes | Experiments were run on an XPS 13 with an Intel Core i7. |
| Software Dependencies | No | The paper does not mention any specific software dependencies with version numbers. It mentions implementing Algorithm 4 but does not specify the programming language or any libraries used. |
| Experiment Setup | Yes | Our experiments focus on the setting where the prior distribution is a d-dimensional Gaussian that is independent across all dimensions and has mean 0.1 in the first dimension and mean 0 in all other dimensions. Note that our algorithm can be applied to arbitrary prior distributions, however we ran experiments for the independent case as this simplifies the code significantly. For this setting, Assumption 3 holds for ϵd = 0.1, cd = 1, K = 1, and cv = 1. |