AR
Andrej Risteski
cs.LGstat.MLcs.DScs.CLcs.AImath.PRcs.NAmath.NAmath.STstat.TH
On Valency
published · living versionsW_xmfazpnx·v1 · currentpublished
Understanding the Benefits of Scaffolds through Computational Separations
with Jingchu Gai, Adam Block
1 version
Preprints & journals
66 papers in the corpus · 2012–2026Understanding the Benefits of Scaffolds through Computational Separations2610.27157v1 · Jingchu Gai, Adam Block, Andrej Risteski2026 · 0 citationsarXivon Valency
When Does Online Imitation Learning Help in LLM Post-Training? The Role of (Non-)Realizability Beyond Horizon2606.30445v1 · Huaqing Zhang, Jingchu Gai, Juno Kim et al.2026 · 0 citationsarXiv
Understanding and Mitigating Premature Confidence for Better LLM Reasoning2605.24396v2 · Jingchu Gai, Guanning Zeng, Christina Baek et al.2026 · 0 citationsarXiv
A computational phase transition for learning-to-sample from Ising models2605.24752v1 · Andrej Risteski, Thuy-Duong Vuong2026 · 0 citationsarXiv
The tractability landscape of diffusion alignment: regularization, rewards, and computational primitives2605.11361v1 · Ankur Moitra, Andrej Risteski, Dhruv Rohatgi2026 · 0 citationsarXiv
Steering diffusion models with quadratic rewards: a fine-grained analysis2602.16570v1 · Ankur Moitra, Andrej Risteski, Dhruv Rohatgi2026 · 0 citationsarXiv
Toward Understanding the Transferability of Adversarial Suffixes in Large Language Models2510.22014v2 · Sarah Ball, Niki Hasrati, Alexander Robey et al.2025 · 0 citationsarXiv
Taming Imperfect Process Verifiers: A Sampling Perspective on Backtracking2510.03149v1 · Dhruv Rohatgi, Abhishek Shetty, Donya Saless et al.2025 · 0 citationsarXiv
CodePDE: An Inference Framework for LLM-driven PDE Solver Generation2505.08783v2 · Shanda Li, Tanya Marwah, Junhong Shen et al.2025 · 0 citationsarXiv
On the Query Complexity of Verifier-Assisted Language Generation2502.12123v2 · Edoardo Botta, Yuchen Li, Aashay Mehta et al.2025 · 0 citationsarXiv
Towards characterizing the value of edge embeddings in Graph Neural Networks2410.09867v1 · Dhruv Rohatgi, Tanya Marwah, Zachary Chase Lipton et al.2024 · 0 citationsarXiv
Progressive distillation induces an implicit curriculum2410.05464v1 · Abhishek Panigrahi, Bingbin Liu, Sadhika Malladi et al.2024 · 0 citationsarXiv
On the Benefits of Memory for Modeling Time-Dependent PDEs2409.02313v2 · Ricardo Buitrago Ruiz, Tanya Marwah, Albert Gu et al.2024 · 1 citationarXiv
Promises and Pitfalls of Generative Masked Language Modeling: Theoretical Framework and Practical Guidelines2407.21046v1 · Yuchen Li, Alexandre Kirchmeyer, Aashay Mehta et al.2024 · 0 citationsarXiv
Transformers are uninterpretable with myopic methods: a case study with bounded Dyck grammars2312.01429v1 · Kaiyue Wen, Yuchen Li, Bingbin Liu et al.2023 · 0 citationsarXiv
Deep Equilibrium Based Neural Operators for Steady-State PDEs2312.00234v1 · Tanya Marwah, Ashwini Pokle, J. Zico Kolter et al.2023 · 6 citationsarXiv
Outliers with Opposing Signals Have an Outsized Effect on Neural Network Optimization2311.04163v1 · Elan Rosenfeld, Andrej Risteski2023 · 0 citationsarXiv
Provable benefits of annealing for estimating normalizing constants: Importance Sampling, Noise-Contrastive Estimation, and beyond2310.03902v2 · Omar Chehab, Aapo Hyvarinen, Andrej Risteski2023 · 1 citationarXiv
Fit Like You Sample: Sample-Efficient Generalized Score Matching from Fast Mixing Diffusions2306.09332v3 · Yilong Qin, Andrej Risteski2023 · 0 citationsarXiv
Provable benefits of score matching2306.01993v1 · Chirag Pabbaraju, Dhruv Rohatgi, Anish Sevekari et al.2023 · 1 citationarXiv
Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression2306.00788v3 · Runtian Zhai, Bingbin Liu, Andrej Risteski et al.2023 · 2 citationsarXiv
How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding2303.04245v2 · Yuchen Li, Yuanzhi Li, Andrej Risteski2023 · 4 citationsarXiv
Neural Network Approximations of PDEs Beyond Linearity: A Representational Perspective2210.12101v2 · Tanya Marwah, Zachary C. Lipton, Jianfeng Lu et al.2022 · 2 citationsarXiv
Statistical Efficiency of Score Matching: The View from Isoperimetry2210.00726v2 · Frederic Koehler, Alexander Heckett, Andrej Risteski2022 · 8 citationsarXiv
Pitfalls of Gaussians as a noise distribution in NCE2210.00189v2 · Holden Lee, Chirag Pabbaraju, Anish Sevekari et al.2022 · 0 citationsarXiv
Contrasting the landscape of contrastive and non-contrastive learning2203.15702v1 · Ashwini Pokle, Jinjin Tian, Yuchen Li et al.2022 · 2 citationsarXiv
Continual learning: a feature extraction formalization, an efficient algorithm, and fundamental obstructions2203.14383v1 · Binghui Peng, Andrej Risteski2022 · 2 citationsarXiv
Masked prediction tasks: a parameter identifiability view2202.09305v1 · Bingbin Liu, Daniel Hsu, Pradeep Ravikumar et al.2022 · 0 citationsarXiv
Sampling Approximately Low-Rank Ising Models: MCMC meets Variational Methods2202.08907v1 · Frederic Koehler, Holden Lee, Andrej Risteski2022 · 2 citationsarXiv
Domain-Adjusted Regression or: ERM May Already Learn Features Sufficient for Out-of-Distribution Generalization2202.06856v2 · Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski2022 · 9 citationsarXiv
Variational autoencoders in the presence of low-dimensional data: landscape and implicit bias2112.06868v2 · Frederic Koehler, Viraj Mehta, Chenghui Zhou et al.2021 · 0 citationsarXiv
Analyzing and Improving the Optimization Landscape of Noise-Contrastive Estimation2110.11271v1 · Bingbin Liu, Elan Rosenfeld, Pradeep Ravikumar et al.2021 · 0 citationsarXiv
The Effects of Invertibility on the Representational Complexity of Encoders in Variational Autoencoders2107.04652v1 · Divyansh Pareek, Andrej Risteski2021 · 0 citationsarXiv
Universal Approximation for Log-concave Distributions using Well-conditioned Normalizing Flows2107.02951v1 · Holden Lee, Chirag Pabbaraju, Anish Sevekari et al.2021 · 10 citationsarXiv
Iterative Feature Matching: Toward Provable Domain Generalization with Logarithmic Environments2106.09913v2 · Yining Chen, Elan Rosenfeld, Mark Sellke et al.2021 · 7 citationsarXiv
The Limitations of Limited Context for Constituency Parsing2106.01580v1 · Yuchen Li, Andrej Risteski2021 · 0 citationsarXiv
Parametric Complexity Bounds for Approximating PDEs with Neural Networks2103.02138v2 · Tanya Marwah, Zachary C. Lipton, Andrej Risteski2021 · 6 citationsarXiv
Contrastive learning of strong-mixing continuous-time stochastic processes2103.02740v1 · Bingbin Liu, Pradeep Ravikumar, Andrej Risteski2021 · 0 citationsarXiv
An Online Learning Approach to Interpolation and Extrapolation in Domain Generalization2102.13128v2 · Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski2021 · 8 citationsarXiv
The Risks of Invariant Risk Minimization2010.05761v2 · Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski2020 · 103 citationsarXiv
Representational aspects of depth and conditioning in normalizing flows2010.01155v2 · Frederic Koehler, Viraj Mehta, Andrej Risteski2020 · 10 citationsarXiv
Efficient sampling from the Bingham distribution2010.00137v2 · Rong Ge, Holden Lee, Jianfeng Lu et al.2020 · 3 citationsAlgorithmic Learning Theory. PMLR, 2021
On Learning Language-Invariant Representations for Universal Machine Translation2008.04510v1 · Han Zhao, Junjie Hu, Andrej Risteski2020 · 4 citationsarXiv
Fast Convergence for Langevin Diffusion with Manifold Structure2002.05576v2 · Ankur Moitra, Andrej Risteski2020 · 6 citationsarXiv
Empirical Study of the Benefits of Overparameterization in Learning Latent Variable Models1907.00030v3 · Rares-Darius Buhai, Yoni Halpern, Yoon Kim et al.2019 · 19 citationsarXiv
Sum-of-squares meets square loss: Fast rates for agnostic tensor completion1905.13283v1 · Dylan J. Foster, Andrej Risteski2019 · 1 citationarXiv
Simulated Tempering Langevin Monte Carlo II: An Improved Proof using Soft Markov Chain Decomposition1812.00793v3 · Rong Ge, Holden Lee, Andrej Risteski2018 · 10 citationsAdvances in Neural Information Processing Systems 31 (2018)
Mean-field approximation, convex hierarchies, and the optimality of correlation rounding: a unified perspective1808.07226v2 · Vishesh Jain, Frederic Koehler, Andrej Risteski2018 · 23 citationsarXiv
Approximability of Discriminators Implies Diversity in GANs1806.10586v4 · Yu Bai, Tengyu Ma, Andrej Risteski2018 · 48 citationsarXiv
Representational Power of ReLU Networks and Polynomial Kernels: Beyond Worst-Case Analysis1805.11405v1 · Frederic Koehler, Andrej Risteski2018 · 5 citationsarXiv
Career total: 83 works. 66 are in this corpus.Showing the 50 most recent.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.