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..
Non-equilibrium Annealed Adjoint Sampler
Authors: Jaemoo Choi, Yongxin Chen, Molei Tao, Guan-Horng Liu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of NAAS across a range of tasks, including sampling from classical energy landscapes and molecular Boltzmann distributions. [...] We evaluate the effectiveness of NAAS on standard synthetic benchmarks and a challenging molecular generation task involving alanine dipeptide, demonstrating stable convergence and strong sample quality across settings. [...] 5 Experiments In this section, we evaluate NAAS across a diverse set of sampling benchmarks. [...] Table 2: Results on synthetic energy functions. We report Sinkhorn distance ( ) and MMD ( ). |
| Researcher Affiliation | Collaboration | 1Georgia Institute of Technology, 2FAIR at Meta, Core contributors EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1 Non-equilibrium Annealed Adjoint Sampler (NAAS) |
| Open Source Code | No | Due to organizational policy, we are currently unable to release our source code, but we plan to make it publicly available as soon as possible. |
| Open Datasets | Yes | Our primary focus on challenging synthetic distributions, including the 32-mode many-well (MW54, 5d) distribution, the 10d Funnel benchmark, a 40-mode Gaussian mixture model (GMM40) in 50d, and a 50d Student mixture model (Mo S). [...] Following the implementation of Midgley et al. (2023); Wu et al. (2020), we use the Open MM library (Eastman et al., 2017) and represent the molecular structure using internal coordinates, resulting in a 60-dimensional state space. [...] All data used in this work is open-source. |
| Dataset Splits | No | The paper uses standard synthetic benchmarks and a molecular generation task. It mentions using '2000 generated and reference samples' for Alanine Dipeptide, but does not specify training, test, or validation splits for any of the datasets used in the experiments. It refers to 'standard synthetic benchmarks' but does not provide details on how the data was partitioned for these experiments. |
| Hardware Specification | Yes | We also report wall-clock time for the 50D setting, measured on a single NVIDIA A6000 GPU. |
| Software Dependencies | No | The paper mentions using 'Open MM library (Eastman et al., 2017)' and 'Python OT (POT) library (Flamary et al., 2021)' and 'Adam optimizer (Kingma and Ba, 2014)'. However, it does not provide specific version numbers for the Open MM or Python OT libraries, nor does it specify a version for the Adam optimizer (which is an algorithm, not a software library itself). |
| Experiment Setup | Yes | Table 12: Hyperparameter Settings [...] The hyperparameters for each experiment are organized in Table 12. |