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 the Mechanisms of Weak-to-Strong Generalization: A Theoretical Perspective
Authors: Behrad Moniri, Hamed Hassani
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | In this paper, through a theoretical analysis of simple models, we uncover three core mechanisms that can drive this phenomenon. ... Although our results are proven for this asymptotic regime, through numerical experiments, we show that they still match simulations very well, even for moderately large values of ns, nt, d X. ... The paper explores theoretical models and uses numerical simulations to validate theoretical predictions rather than conducting empirical studies on real-world datasets for performance evaluation. |
| Researcher Affiliation | Academia | Behrad Moniri University of Pennsylvania Philadelphia, PA EMAIL Hamed Hassani University of Pennsylvania Philadelphia, PA EMAIL |
| Pseudocode | No | The paper does not contain explicit pseudocode or algorithm blocks. It describes mathematical models and derivations. |
| Open Source Code | No | The paper does not contain any statement about releasing source code or provide a link to a code repository. |
| Open Datasets | No | The paper uses synthetically generated data based on theoretical distributions (e.g., "xi N(0, Id X), yi = β xi + εi (1)" and multi-index functions) rather than existing public datasets. There is no concrete access information for any public dataset. |
| Dataset Splits | No | The paper describes generating 'nt independent samples St' and 'ns N unlabeled covariates Ss' from theoretical distributions for simulations, rather than using predefined splits of an existing dataset. Specific train/test/validation splits in the conventional sense are not provided. |
| Hardware Specification | No | The paper mentions 'numerical simulations' but does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run these simulations. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers, such as programming languages, libraries, or solvers, that would be needed to replicate the experiments. |
| Experiment Setup | Yes | In Section 4 'Numerical Validation', the paper specifies parameters used for its simulations, such as 'd X = 500, nt = ns = 2000, σε = 1', 'ζ = 0.8', 'λt = σ2 εγt = 0.25'. These details act as the experimental setup parameters for their numerical validations. |