publications

2026

1. Preprint

   Unlocking Lossless Speedups in LLMs via Discrete Diffusion

   Subham Sekhar Sahoo^*, Lingjie Chen^*, Khiem Pham^*, Jonathan Geuter^*, Chaitanya Dwivedi, Varad Pimpalkhute, Yash Akhauri, Alexander Moreno, and 9 more authors

   arXiv preprint arXiv:2609.04010, 2026

   URL Bib PDF

   @article{sahoo2026unlocking,
     title = {{U}nlocking {L}ossless {S}peedups in {LLM}s via {D}iscrete {D}iffusion},
     author = {Sahoo, Subham Sekhar and Chen, Lingjie and Pham, Khiem and Geuter, Jonathan and Dwivedi, Chaitanya and Pimpalkhute, Varad and Akhauri, Yash and Moreno, Alexander and Yurochkin, Mikhail and Wang, Zhenting and Elhoushi, Mostafa and Dey, Nolan and Bergsma, Shane and Hestness, Joel and Thickstun, John and Xing, Eric and Liu, Zhengzhong},
     journal = {arXiv preprint arXiv:2609.04010},
     year = {2026},
     url = {https://arxiv.org/abs/2609.04010},
   }

2. EMNLP

   Thinking at the Right Size: Amortized Distillation Across Post-Trained LLMs

   Yan Zhou, Sara Kangaslahti^*, Jonathan Geuter^*, Nihal V. Nayak, Marco Fumero, Francesco Locatello, and David Alvarez-Melis

   Findings of the Association for Computational Linguistics: EMNLP 2026, Budapest, Hungary , 2026

   URL Bib PDF

   @inproceedings{zhou2026thinking,
     title = {{T}hinking at the {R}ight {S}ize: {A}mortized {D}istillation {A}cross {P}ost-{T}rained {LLM}s},
     author = {Zhou, Yan and Kangaslahti, Sara and Geuter, Jonathan and Nayak, Nihal V. and Fumero, Marco and Locatello, Francesco and Alvarez-Melis, David},
     booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2026, Budapest, Hungary},
     year = {2026},
     url = {https://arxiv.org/abs/2608.22854},
   }

3. NeurIPS

   Understanding Layer Patching in Model Size Interpolation

   Sara Kangaslahti, Jonathan Geuter^*, Nihal V. Nayak^*, Marco Fumero^*, Francesco Locatello, and David Alvarez-Melis

   Advances in Neural Information Processing Systems (NeurIPS) , 2026

   Previously accepted at AdaptFM: Resource-Adaptive Foundation Model Inference Workshop at ICML 2026

   URL Bib PDF

   @inproceedings{kangaslahti2026understanding,
     title = {{U}nderstanding {L}ayer {P}atching in {M}odel {S}ize {I}nterpolation},
     author = {Kangaslahti, Sara and Geuter, Jonathan and Nayak, Nihal V. and Fumero, Marco and Locatello, Francesco and Alvarez-Melis, David},
     booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
     year = {2026},
     url = {https://arxiv.org/abs/2607.08170},
     note = {Previously accepted at AdaptFM: Resource-Adaptive Foundation Model Inference Workshop at ICML 2026}
   }

4. ICML

   Stop Training for the Worst: Progressive Unmasking Accelerates Masked Diffusion Training

   Jaeyeon Kim^*, Jonathan Geuter^*, David Alvarez-Melis, Sham Kakade, and Sitan Chen

   Forty-third International Conference on Machine Learning (ICML) , 2026

   Previously accepted at MM Intelligence Workshop at ICLR 2026

   URL Bib PDF

   @inproceedings{kim2026stoptrainingworstprogressive,
     title = {{S}top {T}raining for the {W}orst: {P}rogressive {U}nmasking {A}ccelerates {M}asked {D}iffusion {T}raining},
     author = {Kim, Jaeyeon and Geuter, Jonathan and Alvarez-Melis, David and Kakade, Sham and Chen, Sitan},
     booktitle = {Forty-third International Conference on Machine Learning (ICML)},
     year = {2026},
     url = {https://arxiv.org/abs/2602.10314},
     note = {Previously accepted at MM Intelligence Workshop at ICLR 2026}
   }

5. ICLR

   Boomerang Distillation Enables Zero-Shot Model Size Interpolation

   Sara Kangaslahti, Nihal V. Nayak^*, Jonathan Geuter^*, Marco Fumero, Francesco Locatello, and David Alvarez-Melis

   Proceedings of the International Conference on Learning Representations (ICLR) , 2026

   Previously accepted at 3rd Workshop on Unifying Representations in Neural Models at NeurIPS 2025

   URL Bib PDF

   @inproceedings{kangaslahti2025boomerang,
     author = {Kangaslahti, Sara and Nayak, Nihal V. and Geuter, Jonathan and Fumero, Marco and Locatello, Francesco and Alvarez-Melis, David},
     title = {{B}oomerang {D}istillation {E}nables {Z}ero-{S}hot {M}odel {S}ize {I}nterpolation},
     booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)},
     url = {https://arxiv.org/abs/2510.05064},
     year = {2026},
     note = {Previously accepted at 3rd Workshop on Unifying Representations in Neural Models at NeurIPS 2025}
   }

6. ICLR

   Guided Speculative Inference for Efficient Test-Time Alignment of LLMs

   Jonathan Geuter, Youssef Mroueh, and David Alvarez-Melis

   Proceedings of the International Conference on Learning Representations (ICLR) , 2026

   Previously accepted at 3rd Workshop for Efficient Systems for Foundation Models at ICML 2025 (Spotlight)

   URL Bib PDF

   @inproceedings{Geuter2025GuidedSpeculative,
     author = {Geuter, Jonathan and Mroueh, Youssef and Alvarez-Melis, David},
     title = {{G}uided {S}peculative {I}nference for {E}fficient {T}est-{T}ime {A}lignment of {LLM}s},
     booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)},
     url = {https://arxiv.org/abs/2506.04118},
     year = {2026},
     note = {Previously accepted at 3rd Workshop for Efficient Systems for Foundation Models at ICML 2025 (Spotlight)},
   }

2025

1. NeurIPS (Workshop)

   RoBoN: Routed Online Best-of-n for Test-Time Scaling with Multiple LLMs

   Jonathan Geuter and Gregor Kornhardt

   1st Workshop on Foundations of Reasoning in Language Models at NeurIPS 2025, San Diego, USA , 2025

   URL Bib PDF

   @inproceedings{geuter2025robon,
     author = {Geuter, Jonathan and Kornhardt, Gregor},
     title = {{R}o{B}o{N}: {R}outed {O}nline {B}est-of-{n} for {T}est-{T}ime {S}caling with {M}ultiple {LLM}s},
     booktitle = {1st Workshop on Foundations of Reasoning in Language Models at NeurIPS 2025, San Diego, USA},
     url = {https://arxiv.org/abs/2512.05542},
     year = {2025}
   }

2. ICML (Workshop)

   Entropy-Driven Pre-Tokenization for Byte-Pair Encoding

   Yifan Hu, Frank Liang, Dachuan Zhao, Jonathan Geuter, Varshini Reddy, Craig W. Schmidt, and Chris Tanner

   Tokenization Workshop at ICML , 2025

   URL Bib PDF

   @inproceedings{hu2025entropydriven,
     title = {{E}ntropy-{D}riven {P}re-{T}okenization for {B}yte-{P}air {E}ncoding},
     author = {Hu, Yifan and Liang, Frank and Zhao, Dachuan and Geuter, Jonathan and Reddy, Varshini and Schmidt, Craig W. and Tanner, Chris},
     booktitle = {Tokenization Workshop at ICML},
     year = {2025},
     url = {https://openreview.net/forum?id=1hcu9Vq4Or},
   }

3. AISTATS

   DDEQs: Distributional Deep Equilibrium Models through Wasserstein Gradient Flows

   Jonathan Geuter, Clément Bonet, Anna Korba, and David Alvarez-Melis

   The 28th International Conference on Artificial Intelligence and Statistics (AISTATS) , 2025

   URL Bib PDF

   @inproceedings{geuter2025ddeqs,
     title = {{DDEQ}s: {D}istributional {D}eep {E}quilibrium {M}odels through {W}asserstein {G}radient {F}lows},
     author = {Geuter, Jonathan and Bonet, Cl{\'e}ment and Korba, Anna and Alvarez-Melis, David},
     booktitle = {The 28th International Conference on Artificial Intelligence and Statistics (AISTATS)},
     year = {2025},
     url = {https://openreview.net/forum?id=rFfNuzzXXW},
   }

4. ICML

   Universal Neural Optimal Transport

   Jonathan Geuter, Gregor Kornhardt, Ingimar Tomasson, and Vaios Laschos

   Forty-second International Conference on Machine Learning (ICML) , 2025

   URL Bib PDF

   @inproceedings{geuter2025universal,
     title = {{U}niversal {N}eural {O}ptimal {T}ransport},
     author = {Geuter, Jonathan and Kornhardt, Gregor and Tomasson, Ingimar and Laschos, Vaios},
     booktitle = {Forty-second International Conference on Machine Learning (ICML)},
     year = {2025},
     url = {https://openreview.net/forum?id=t10fde8tQ7},
   }

2023

1. EMNLP (Workshop)

   Jina Embeddings: A Novel Set of High-Performance Sentence Embedding Models

   Michael Günther, Louis Milliken, Jonathan Geuter, Georgios Mastrapas, Bo Wang, and Han Xiao

   3rd Workshop for Natural Language Processing Open Source Software at EMNLP , 2023

   URL Bib PDF

   @inproceedings{guenther2023jina,
     title = {{J}ina {E}mbeddings: {A} {N}ovel {S}et of {H}igh-{P}erformance {S}entence {E}mbedding {M}odels},
     author = {Günther, Michael and Milliken, Louis and Geuter, Jonathan and Mastrapas, Georgios and Wang, Bo and Xiao, Han},
     booktitle = {3rd Workshop for Natural Language Processing Open Source Software at EMNLP},
     year = {2023},
     url = {https://openreview.net/forum?id=3HTUlrn6o7},
   }

2022

1. A Sinkhorn-NN Hybrid Algorithm for Optimal Transport

   Jonathan Geuter

   Technische Universität Berlin and Weierstrass Institute.Master’s thesis, supervised by Dr. Vaios Laschos and Prof. Dr. Martin Skutella , 2022

   Bib PDF

   @mastersthesis{Geuter2022SinkhornNN,
     author = {Geuter, Jonathan},
     title = {{A} {S}inkhorn-{NN} {H}ybrid {A}lgorithm for {O}ptimal {T}ransport},
     school = {Technische Universität Berlin and Weierstrass Institute.},
     year = {2022},
   }

2020

1. Nonlinear Korn Inequalities

   Jonathan Geuter

   Bachelor’s thesis, supervised by Prof. Dr. Barbara Zwicknagl and Dr. Hans-Christian Kreusler , 2020

   Bib PDF

   @techreport{Geuter2020NonlinearKorn,
     author = {Geuter, Jonathan},
     title = {{N}onlinear {K}orn {I}nequalities},
     institution = {Technische Universität Berlin and Humboldt Universität zu Berlin},
     year = {2020},
   }