Good Papers: read the good ones

A stopgrad regression principle characterizes stationary points of stopgrad objectives and proves convergence to true flow maps while halving training memory.

Mark Goldstein, Max Shen, Zichu Wang, Aahlad Manas Puli and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

83%Must read

?Must readVote to see the score

NeurIPS 2026Georgia Institute of Technology Deep learning theory

SGD Provably Prioritizes a Shortcut Spurious Feature in the XOR Model

SGD learns linear spurious correlations exponentially faster than XOR signals in two-layer ReLU networks, with dynamics that suppress true feature learning.

Tyler LaBonte, Vidya Muthukumar

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026U Central Florida Deep learning theory

Bifurcation Models: Learning Set-Valued Solution Maps with Weight-Tied Dynamics

Bifurcation models learn set-valued solution maps via weight-tied dynamics, representing multi-branch attractor landscapes with almost everywhere regular selectors and outperforming single-branch supervision, though diversity requires explicit encouragement.

Caleb Jore, Jialin Liu

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Imperial College London Universität St. Gallen Deep learning theory

Random Neural Network Expressivity for Non-Linear Partial Differential Equations

Random neural networks approximate time-dependent Sobolev functions with dimension-free rate 1/2 and efficiently solve nonlinear porous medium and compressible Navier-Stokes equations.

Muhammed Ali Mehmood, Lukas Gonon

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

6/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

70%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Stony Brook Deep learning theory

Subcritical Signal Propagation at Initialization in Normalization-Free Transformers

Average partial Jacobian norms in transformers reveal subcritical signal growth in normalization-free architectures via tanh-like nonlinearities, explaining initialization sensitivity in DyT and Derf models.

Sergey Alekseev

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

5/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 5 of 20 reviewers recommend it

lenient 2/5

medium 3/10

strict 0/5

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026SpotlightCopenhagen U Copenhagen U Technology Nuremberg Worcester Polytechnic Institute Technion - Israel institue of Te Deep learning theory

On the Depth of Monotone ReLU Neural Networks and ICNNs

Monotone ReLU networks cannot compute or approximate maximum, ICNNs need depth n for it, and depth-k ICNNs cannot simulate some depth-2 ReLU networks.

Egor Bakaev, Florestan Brunck, Christoph Hertrich, Daniel Reichman and 1 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026SpotlightU Pennsylvania Sakana AI Northeastern Harvard Deep learning theory

Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves

A Hilbert bundle convolutional framework defines HilbNets for infinite-dimensional manifold signals, proving discrete versions converge to continuous architectures and transfer across samplings.

Kartik Tandon, Julian J Gould, Tanishq Bhatia, Francesca Dominici and 2 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 8 of 20 reviewers recommend it

lenient 3/5

medium 3/10

strict 2/5

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT EPFL Princeton EPFL - EPF Lausanne Deep learning theory

Provable Quantization with Randomized Hadamard Transform

Dithered randomized Hadamard quantization is unbiased and achieves mean squared error asymptotically matching dense random rotations at O(d log d) cost.

Ying Feng, Piotr Indyk, Michael Kapralov, Dmitrii Krachun and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

70%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Georgia Institute of Technology U Washington U Illinois at Urbana-Champaign Deep learning theory

Global Convergence of Four-Layer Matrix Factorization under Random Initialization

Gradient descent globally converges for randomly initialized four-layer matrix factorization with balanced regularization, avoiding saddles in polynomial time.

Minrui Luo, Weihang Xu, Xiang Gao, Maryam Fazel and 1 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

5/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Yale Deep learning theory

Decoupled Descent: Exact Test Error Tracking Via Approximate Message Passing

Decoupled descent cancels data-reuse biases via approximate message passing so training error tracks test error, enabling zero-cost validation and shrinking the generalization gap versus gradient descent.

Max Lovig

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026National University of Singapore Deep learning theory

InfoFlow: A Framework for Multi-Layer Transformer Analysis

InfoFlow proves multi-layer Transformers exponentially beat single-layer ones on retrieval tasks and tracks information propagation to explain multi-layer approximation efficiency.

Penghao Yu, Haotian Jiang, Zeyu Bao, Qianxiao Li

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026OralTechnion Technion - Israeli institute of Deep learning theory

Learning Reveals Invisible Structure in Low-Rank RNNs

Deriving reduced ODEs for low-rank RNN learning reveals loss-invisible overlaps that encode training history and expose hidden connectivity differences.

Yoav Ger, Omri Barak

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Oxford Boston Deep learning theory

Convergence Analysis of Newton's Method for Neural Networks in the Overparameterized Limit

Regularized Newton training of overparameterized neural networks converges to a deterministic NNTK limit with exponentially fast uniform convergence across all frequencies, avoiding gradient descent's spectral bias.

Konstantin Riedl, Justin Sirignano, Konstantinos Spiliopoulos

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026RIKEN Imperial College London The University of Tokyo / RIKEN Deep learning theory

Training-Induced Escape from Token Clustering in a Mean-Field Formulation of Transformers

Training a linear FFN in mean-field transformers drives token distributions to escape attention-induced clustering near final layers.

Noboru Isobe, Daisuke Inoue, Masaaki Imaizumi

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 8 of 20 reviewers recommend it

lenient 2/5

medium 4/10

strict 2/5

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Friedrich-Alexander Universität U Wuerzburg U Oxford Technical University of Munich Deep learning theory

Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime

Mean-field transformers exhibit rapid token distribution concentration onto projection-driven limits with explicit Wasserstein bounds scaling in inverse temperature β and time t.

Albert Alcalde, Leon Bungert, Konstantin Riedl, Tim Roith

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Gatsby Computational Neuroscienc University College London, Unive IST Austria Stanford Deep learning theory

Optimal Representation Size: High-Dimensional Analysis of Pretraining and Linear Probing

High-dimensional analysis of pretraining via PCA and linear probing derives exact errors versus representation size, showing compression helps with abundant unlabeled but scarce labeled data.

Valentina Njaradi, Clémentine Dominé, Rachel A Swanson, Marco Mondelli and 1 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Columbia Deep learning theory

Fixed Universal Transformers

Fixed universal transformers simulate any target transformer via input embeddings with frozen internal parameters, and random initialization achieves universality almost surely.

Jingwen Liu, Alexandr Andoni, Daniel Hsu

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

67%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Champalimaud Foundation Champalimaud Centre for the Unkn Deep learning theory

Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees

Framework guarantees universal approximation for multistable dynamics with infinite-time horizon guarantees, linking topological properties to training metrics.

Ábel Ságodi, Memming Park

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

2/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 2 of 20 reviewers recommend it

lenient 0/5

medium 2/10

strict 0/5

45%Niche pick

?Niche pickVote to see the score

NeurIPS 2026U Tübingen & IMPRS-IS Allen Institute, Neural Dynamics Stanford U Alberta Deep learning theory

Parameter symmetries determine representational geometry in overparameterized nonlinear networks

The poster argues parameter symmetries determine representational geometry in overparameterized nonlinear networks.

Marvin Theiss, Lukas Braun, Andrew Saxe, Erin Grant

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

0/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 0 of 20 reviewers recommend it

lenient 0/5

medium 0/10

strict 0/5

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026CNRS and ENS Deep learning theory

Muon Dynamics as a Spectral Wasserstein Flow

Spectral Wasserstein distances unify normalized matrix flows, proving Muon dynamics are gradient flows with Benamou-Brenier equivalence for monotone norms.

Gabriel Peyré

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

6/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Sissa SISSA, Trieste, Italy INRIA Deep learning theory

Factual recall in linear associative memories: sharp asymptotics and mechanistic insights

Linear associative memory stores up to ~d² log p / 2 facts by raising correct scores above competing extremes, not via Hebbian broad fluctuations.

Alessio Giorlandino, Sebastian Goldt, Antoine Maillard

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

88%Must read

?Must readVote to see the score

NeurIPS 2026U Texas at Austin Harvard Deep learning theory

Theory of Optimal Learning Rate Schedules and Scaling Laws for a Random Feature Model

Optimal LR schedules for a solvable random feature model reveal easy-phase polynomial decay and hard-phase warmup-stable-decay regimes that improve scaling over constant or power-law schedules, with momentum and batch ramps further enhancing wall-clock time.

Blake Bordelon, Francesco Mori

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

15/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 15 of 20 reviewers recommend it

lenient 4/5

medium 9/10

strict 2/5

83%Must read

?Must readVote to see the score

NeurIPS 2026U Cambridge Deep learning theory

Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors

Score errors decompose into visible gradient and invisible solenoidal parts, so L2 score error cannot bound distribution divergence and only gradient error matters for diffusion sampling quality.

Nail B Khelifa, Richard Turner, Ramji Venkataramanan

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

70%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Stuttgart Aarhus Deep learning theory

Statistical Convergence of Spherical First Hitting Diffusion Models

Spherical first-hitting diffusion models achieve near-minimax optimal convergence rates in total variation for Sobolev data on spheres, marking the first statistical optimality result for diffusion models using random generation times.

Simon Bienewald, Lukas Trottner

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

5/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026 Max Planck Institute for Mathem UCLA Deep learning theory

Most ReLU Networks Admit Identifiable Parameters

Deep ReLU networks with input and hidden widths ≥2 have open sets of identifiable parameters, yielding exact functional dimensions and generic depth hierarchies.

Moritz Grillo, Guido Montufar

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 7 of 20 reviewers recommend it

lenient 1/5

medium 4/10

strict 2/5

70%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026SpotlightTechnical University of Denmark Max Planck Institute for Mathem UCLA Deep learning theory

The Symmetries of Three-Layer ReLU Networks

This paper characterizes generic parameter fibers of three-layer ReLU bottleneck networks via semi-algebraic descriptions, yielding polynomial-time functional equivalence tests and analyzing symmetry-induced gradient flow conservation laws.

Johanna Marie Gegenfurtner, Moritz Grillo, Guido Montufar

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

5/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 5 of 20 reviewers recommend it

lenient 1/5

medium 3/10

strict 1/5

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Doshisha Rice U Basel U Helsinki Deep learning theory

Function graph transformers universally approximate operators between function spaces

Function graph transformers lift functions to graph measures to universally approximate nonlinear operators between function spaces via standard attention and MLPs.

Takashi Furuya, S D Mis, Ivan Dokmanić, Maarten V. de Hoop and 1 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Massachusetts Institute of Techn Deep learning theory

Training with (Swap) Regret Loss in a Single-Layer Self-Attention Model: A Case Study on the Probability Simplex

Single-layer self-attention trained with regret and swap-regret loss learns smoothed fictitious play and Blum-Mansour dynamics yielding coarse and correlated equilibria without supervised traces.

Chanwoo Park, Asuman Ozdaglar

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Alignment Research Center Stanford OpenAI Deep learning theory

Estimating the expected output of wide random MLPs more efficiently than sampling

Approximate layer-wise activation distributions via cumulants and Hermite expansions to estimate wide MLP expected outputs without sampling, reducing FLOPs versus Monte Carlo and improving rare-event estimates.

Wilson Wu, Victor Lecomte, Michael Winer, George Robinson and 2 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MPI-SWS Rheinland-Pfälzische Technische U Notre Dame U Kaiserslautern RPTU, University of Kaiserslaute Deep learning theory

Length Generalization for Transformers via Compression

Refining C-RASP via compressed strings yields polynomial transformer length generalization bounds for tasks with compressed-string solutions.

Georg Zetzsche, Hongjian Jiang, Andy J Yang, Pascal Bergsträßer and 3 more

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 2/5

medium 6/10

strict 3/5

70%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026UQAM Deep learning theory

Ordinary Least Squares as an Attention Mechanism

Ordinary least squares predictions are rewritten as restricted attention outputs, framing OLS as similarity-based prediction via learned embedding and decoding operations mapped onto query-key-value structures.

Philippe Goulet Coulombe

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

4/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 4 of 20 reviewers recommend it

lenient 3/5

medium 1/10

strict 0/5

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Duke Deep learning theory

Conservation Laws for Diffusion Models

Conservation laws express diffusion cross-entropy via local information-theoretic derivatives along noise paths, unifying discrete and continuous likelihoods and reducing training to marginal posterior learning.

Ziv Aharoni, Henry Pfister

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026DePaul Iowa State Deep learning theory

Feature Learning Dynamics in Infinite-Depth Neural Networks

For one-layer ResNets under depth-μP scaling, reused-weight forward-backward coupling vanishes at initialization but SGD induces surviving correlations suppressed by depth, yielding a rigorous infinite-depth Neural Feature Dynamics limit.

Zihan Yao, Ruoyu Wu, Tianxiang Gao

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 10 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 3/5

88%Must read

?Must readVote to see the score

NeurIPS 2026Georgia Institute of Technology Georgia Tech Deep learning theory

Bounding Global and Local Compression Error of Signal Parameterizations

A framework predicts reconstruction error of compressive signal parameterizations via scaled differences between model predictions at different compression levels without ground truth. It yields non-asymptotic, signal-specific bounds that closely track global errors and local error heatmaps across i

Quang Luong Nhat Nguyen, Sara Fridovich-Keil

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

15/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026CMU, Carnegie Mellon Carnegie Mellon Massachusetts Institute of Techn Deep learning theory

Concept Modulation Models: A Unified Framework for Identifiability and Extrapolation

Concept modulation models unify conditional latent variable model identifiability and extrapolation via attribute potentials and algebraic criteria for unseen attributes.

Soheun Yi, Yizhou Lu, Chandler Squires, Pradeep Ravikumar

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 8 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 1/5

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Hanoi University of Science and Purdue U Massachusetts at Amherst The University of Texas at Austi U Texas at Austin Deep learning theory

A Statistical Theory of Gated Attention through the Lens of Hierarchical Mixture of Experts

Gated attention represents attention matrices as hierarchical mixtures of experts and achieves polynomial sample complexity versus exponential for multi-head self-attention.

Viet Nguyen, Thinh Cao, Tuan M Pham, Tan Dinh and 3 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Swiss Federal Institute of Techn EPFL EPFL - EPF Lausanne EPFL - Lausanne Deep learning theory

Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning

Neural LoFi frames deep training as iterative spectral low-degree filtering, predicting layer-wise feature selection, concept emergence, and compositional depth via low-degree correlation dynamics.

Yatin Dandi, Matteo Vilucchio, Luca Arnaboldi, Hugo Tabanelli and 1 more

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Washington Deep learning theory

Learning Orthogonal Multi-Index Models Beyond Small Initialization: Incremental Learning, Competitive Dynamics and Symmetry

Under standard initialization, two-layer networks learn orthogonal multi-index targets incrementally via competitive neuron dynamics, with lower-order Hermite components recovered before higher-order directions.

Mo Zhou, Weihang Xu, Simon Du, Maryam Fazel

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 8 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 1/5

88%Must read

?Must readVote to see the score

NeurIPS 2026Brown Deep learning theory

Stability and Generalization in Looped Transformers

A fixed-point framework proves looped transformers need recall plus outer normalization for stable, input-dependent extrapolation, validated across chess, sudoku, and prefix-sums tasks.

Asher Labovich

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

15/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 15 of 20 reviewers recommend it

lenient 4/5

medium 8/10

strict 3/5

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026INRIA Imperial College London Inria Paris / ENS Deep learning theory

Generalization at the Edge of Stability

Stochastic optimizers at the edge of stability converge to low-dimensional fractal attractors, and a sharpness-dimension generalization bound reveals that chaotic training depends on the full Hessian spectrum.

Mario Tuci, Caner Korkmaz, Umut Simsekli, Tolga Birdal

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · Code ★ 10

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 2/5

medium 6/10

strict 3/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026OralMassachusetts Institute of Techn NYU Deep learning theory

Uniform-in-Time Weak Propagation of Chaos in Shallow Neural Networks

Shallow neural networks trained via gradient descent exhibit uniform-in-time weak propagation of chaos, yielding poly(d/ε) neuron and sample complexity when mean-field loss decays faster than t^{-2}.

Margalit Glasgow, Joan Bruna

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 9 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 2/5

Show 20 more papers