Resolution ResearchResearch from Resolution and earlier work from Timaeus.https://resolution.orgPatterning the Toy Model of Superpositionhttps://resolution.org/post/patterning-tmshttps://resolution.org/post/patterning-tmsMax Adam, Jesse Hoogland, Daniel Murfet. Research post.Thu, 29 Oct 2026 00:00:00 GMTFormalizing Hironaka’s Resolution of Singularitieshttps://resolution.org/post/formalizing-resolution-of-singularitieshttps://resolution.org/post/formalizing-resolution-of-singularitiesBilly Snikkers, Adam Newgas, Jesse Hoogland, Chris Elliott. Research post.Tue, 06 Oct 2026 00:00:00 GMTInterpretability as a Benchmark for Samplinghttps://resolution.org/post/interpretability-as-a-benchmark-for-samplinghttps://resolution.org/post/interpretability-as-a-benchmark-for-samplingAndrew Gordon, Rohan Hitchcock, Ioannis Mouratidis, Johan Wind, Daniel Murfet. Research post.Thu, 17 Sep 2026 00:00:00 GMTElicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posteriorhttps://arxiv.org/abs/2607.18804https://arxiv.org/abs/2607.18804Garrett Baker, Vinayak Pathak, Daniel Murfet, Susan Wei. Preprint.Tue, 21 Jul 2026 00:00:00 GMTInterpreting Reinforcement Learning Agents with Susceptibilitieshttps://arxiv.org/abs/2605.08007https://arxiv.org/abs/2605.08007Chris Elliott, Einar Urdshals, David Quarel, Daniel Murfet. Preprint.Fri, 08 May 2026 00:00:00 GMTLinear Response Estimators for Singular Statistical Modelshttps://arxiv.org/abs/2605.07970https://arxiv.org/abs/2605.07970Chris Elliott, Daniel Murfet. Preprint.Fri, 08 May 2026 00:00:00 GMTSusceptibilities and Patterning: A Primer on Linear Response in Bayesian Learninghttps://arxiv.org/abs/2605.07980https://arxiv.org/abs/2605.07980Chris Elliott, Daniel Murfet. Preprint.Fri, 08 May 2026 00:00:00 GMTSpectroscopy at Scale: Finding Interpretable Structure in Pythia-1.4Bhttps://timaeus.co/research/2026-04-21-spectroscopy-mainhttps://timaeus.co/research/2026-04-21-spectroscopy-mainDaniel Murfet, Andrew Gordon, Max Adam, George Wang, Jesse Hoogland, Garrett Baker, William Snell, Stan van Wingerden, Adam Newgas, Billy Snikkers, Rohan Hitchcock. Technical report.Tue, 21 Apr 2026 00:00:00 GMTHow to Scale Susceptibilitieshttps://timaeus.co/research/2026-04-21-spectroscopy-definitionshttps://timaeus.co/research/2026-04-21-spectroscopy-definitionsAndrew Gordon, Max Adam, Jesse Hoogland, Daniel Murfet. Technical report.Tue, 21 Apr 2026 00:00:00 GMTInterpreting the Ising Modelhttps://timaeus.co/research/2026-04-21-spectroscopy-isinghttps://timaeus.co/research/2026-04-21-spectroscopy-isingAndrew Gordon, Rohan Hitchcock, Daniel Murfet. Technical report.Tue, 21 Apr 2026 00:00:00 GMTGuide for Sampling Hyperparameter Selectionhttps://timaeus.co/research/2026-04-21-sampling-guidehttps://timaeus.co/research/2026-04-21-sampling-guideRohan Hitchcock, Jesse Hoogland, George Wang, Andrew Gordon. Technical report.Tue, 21 Apr 2026 00:00:00 GMTPatterning: The Dual of Interpretabilityhttps://arxiv.org/abs/2601.13548https://arxiv.org/abs/2601.13548George Wang, Daniel Murfet. Preprint.Tue, 20 Jan 2026 00:00:00 GMTTowards Spectroscopy: Susceptibility Clusters in Language Modelshttps://arxiv.org/abs/2601.12703https://arxiv.org/abs/2601.12703Andrew Gordon, Garrett Baker, George Wang, William Snell, Stan van Wingerden, Daniel Murfet. Preprint.Mon, 19 Jan 2026 00:00:00 GMTStagewise Reinforcement Learning and the Geometry of the Regret Landscapehttps://arxiv.org/abs/2601.07524https://arxiv.org/abs/2601.07524Chris Elliott, Einar Urdshals, David Quarel, Matthew Farrugia-Roberts, Daniel Murfet. Preprint.Mon, 12 Jan 2026 00:00:00 GMTCompressibility Measures Complexity: Minimum Description Length Meets Singular Learning Theoryhttps://arxiv.org/abs/2510.12077https://arxiv.org/abs/2510.12077Einar Urdshals, Edmund Lau, Jesse Hoogland, Stan van Wingerden, Daniel Murfet. Preprint.Tue, 14 Oct 2025 00:00:00 GMTInfluence Dynamics and Stagewise Data Attributionhttps://arxiv.org/abs/2510.12071https://arxiv.org/abs/2510.12071Jin Hwa Lee, Matthew Smith, Maxwell Adam, Jesse Hoogland. Preprint.Tue, 14 Oct 2025 00:00:00 GMTThe Loss Kernel: A Geometric Probe for Deep Learning Interpretabilityhttps://arxiv.org/abs/2509.26537https://arxiv.org/abs/2509.26537Maxwell Adam, Zach Furman, Jesse Hoogland. Preprint.Wed, 01 Oct 2025 00:00:00 GMTBayesian Influence Functions for Hessian-Free Data Attributionhttps://arxiv.org/abs/2509.26544https://arxiv.org/abs/2509.26544Philipp Alexander Kreer, Wilson Wu, Maxwell Adam, Zach Furman, Jesse Hoogland. Preprint.Tue, 30 Sep 2025 00:00:00 GMTEmbryology of a Language Modelhttps://arxiv.org/abs/2508.00331https://arxiv.org/abs/2508.00331George Wang, Garrett Baker, Andrew Gordon, Daniel Murfet. Preprint.Fri, 01 Aug 2025 00:00:00 GMTFrom Global to Local: A Scalable Benchmark for Local Posterior Samplinghttps://arxiv.org/abs/2507.21449https://arxiv.org/abs/2507.21449Rohan Hitchcock, Jesse Hoogland. Preprint.Tue, 29 Jul 2025 00:00:00 GMTStructural Inference: Interpreting Small Language Models with Susceptibilitieshttps://arxiv.org/abs/2504.18274https://arxiv.org/abs/2504.18274Garrett Baker, George Wang, Jesse Hoogland, Vinayak Pathak, Daniel Murfet. Preprint.Fri, 25 Apr 2025 00:00:00 GMTModes of Sequence Models and Learning Coefficientshttps://arxiv.org/abs/2504.18048https://arxiv.org/abs/2504.18048Zhongtian Chen, Daniel Murfet. Preprint.Fri, 25 Apr 2025 00:00:00 GMTPrograms as Singularitieshttps://arxiv.org/abs/2504.08075https://arxiv.org/abs/2504.08075Daniel Murfet, Will Troiani. Preprint.Thu, 10 Apr 2025 00:00:00 GMTYou Are What You Eat – AI Alignment Requires Understanding How Data Shapes Structure and Generalisationhttps://arxiv.org/abs/2502.05475https://arxiv.org/abs/2502.05475Simon Pepin Lehalleur, Jesse Hoogland, Matthew Farrugia-Roberts, Susan Wei, Alexander Gietelink Oldenziel, George Wang, Stan van Wingerden, Zach Furman, Liam Carroll, Daniel Murfet. Preprint.Sat, 08 Feb 2025 00:00:00 GMTStructure Development in List-Sorting Transformershttps://arxiv.org/abs/2501.18666https://arxiv.org/abs/2501.18666Einar Urdshals, Jasmina Urdshals. ICML 2025 SMUNN Workshop.Thu, 30 Jan 2025 00:00:00 GMTDynamics of Transient Structure in In-Context Linear Regression Transformershttps://arxiv.org/abs/2501.17745https://arxiv.org/abs/2501.17745Liam Carroll, Jesse Hoogland, Matthew Farrugia-Roberts, Daniel Murfet. Preprint.Wed, 29 Jan 2025 00:00:00 GMTDifferentiation and Specialization of Attention Heads via the Refined Local Learning Coefficienthttps://openreview.net/forum?id=SUc1UOWndphttps://openreview.net/forum?id=SUc1UOWndpGeorge Wang, Jesse Hoogland, Stan van Wingerden, Zach Furman, Daniel Murfet. ICLR 2025.Fri, 04 Oct 2024 00:00:00 GMTLoss Landscape Degeneracy and Stagewise Development of Transformershttps://tmlr.infinite-conf.org/paper_pages/45qJyBG8Oj.htmlhttps://tmlr.infinite-conf.org/paper_pages/45qJyBG8Oj.htmlJesse Hoogland, George Wang, Matthew Farrugia-Roberts, Liam Carroll, Susan Wei, Daniel Murfet. TMLR.Sun, 04 Feb 2024 00:00:00 GMTThe Local Learning Coefficient: A Singularity-Aware Complexity Measurehttps://proceedings.mlr.press/v258/lau25a.htmlhttps://proceedings.mlr.press/v258/lau25a.htmlEdmund Lau, Zach Furman, George Wang, Daniel Murfet, Susan Wei. AISTATS 2025.Wed, 23 Aug 2023 00:00:00 GMT