Computation

Jesse Bettencourt, Xindi Wu, Matan Atzmon et al.

For practitioners using diffusion models as frozen teachers in downstream pipelines, CARV reduces computational cost by reusing expensive upstream computations across noise resamples.

William Rudman, Abhishek Divekar, Kanishk Jain et al. · amazon-science

For scientists and researchers needing automated statistical modeling, VESTA addresses the bottleneck of model refinement by enabling active data exploration and tool creation, though the gains are incremental over existing agent-based systems.

Kyurae Kim, Samuel Gruffaz, Ji Won Park et al.

For researchers using Langevin Monte Carlo for sampling, this work extends theoretical guarantees to the overdamped regime, showing the exponential integrator remains stable and effective.

Steve Hanneke, Anay Mehrotra, Grigoris Velegkas et al.

Provides a theoretical characterization and first algorithm for a classic but understudied learning model, clarifying the role of membership queries.

Lifu Wei, Yinuo Ren, Naichen Shi et al.

It provides a computationally efficient and unbiased method for inference-time guidance in diffusion models, addressing the bottleneck of repeated score/gradient evaluations.

Xun Huan, Jayanth Jagalur, Youssef Marzouk

For researchers and practitioners in modeling and prediction across sciences and engineering, this survey provides a comprehensive overview of OED methods and identifies key open problems.

Arthur Kosmala, Stephan Günnemann, Meng Gao et al.

For computational chemists and physicists, LSD addresses the serial bottleneck in molecular dynamics to increase single-system throughput without introducing error.

Deepak John Reji

It provides social scientists with a theory-testing instrument for studying attitude dynamics and polarization, though the demonstration is limited to a single case with no quantitative validation.

Paula Cordero-Encinar, Georgy Tyukin, Andrew B. Duncan

For practitioners of LLM post-training, this method offers a scalable way to obtain uncertainty-aware models that handle conflicting data better than single-model approaches.

Zheng Gai, Li Xincheng, Jiang Wangyingjie et al.

For applied researchers needing to design and analyze inter-rater agreement studies, CATEKAPPA removes the programming barrier by offering a graphical interface, but it is an incremental integration of existing methods.

Ayoub Belhadji, Daniel Sharp, Youssef M. Marzouk

Provides a new paradigm for Bayesian inference by enabling deterministic quadrature via particle systems that avoid mode collapse and scale to high dimensions, addressing limitations of MCMC and variational inference.

Mary Letey, Yue M. Lu, Cengiz Pehlevan et al.

Provides theoretical understanding of how real-world data correlations affect in-context learning, revealing architectural mismatches for practitioners.

Jing Jia, Liyue Shen, Guanyang Wang

For practitioners using diffusion models, this provides a zero-cost way to improve diversity in generated batches, with applications in image generation and editing.

Julian Rodemann, Alexander Marquard, Thomas Augustin et al.

This work provides a deterministic, sampling-free approximation for Bayesian predictive uncertainty that is computationally efficient and applicable to various regression models, offering a practical tool for uncertainty quantification.

Jun Hu

This work provides the first automatic, dimension-aware convergence certificates for learned-transport MCMC, which is significant for researchers and practitioners who rely on MCMC for sampling from complex posterior distributions.

Di Wu, Ling Liang, Haizhao Yang

For practitioners in resource-constrained experimental design, this work provides a more robust and scalable BOED approach that overcomes fundamental limitations of KL-based methods.

Antonio Di Noia, Iuri Macocco, Aldo Glielmo et al. · eth-zurich

This addresses a key challenge in unsupervised learning and feature selection for researchers and practitioners dealing with real-world data affected by noise and manifold curvature.

Christopher Nemeth

This work addresses the challenge of generating realistic hypergraphs, which is important for modeling higher-order interactions in network science, but the improvement is incremental over existing methods.

Dario Draca, Takuo Matsubara, Minh-Ngoc Tran

This work addresses optimization challenges for machine learning models with constrained parameters, such as those in variational inference, by enabling efficient natural gradient descent on manifolds.

Jungang Zou, Alex Ziyu Jiang, Qixuan Chen

This work addresses the coding bottleneck in MCMC workflows for Bayesian practitioners, but the results are preliminary and the system's capability is limited to built-in blocks.