mHC: Manifold-Constrained Hyper-Connections

Manifold-Constrained Hyper-Connections restore identity mappings to hyper-connections via manifold projection, improving training stability, scalability, and efficiency at scale.

Zhenda Xie, Yixuan Wei, Huanqi Cao, Chenggang Zhao and 16 more

Published Dec 31, 2025 · 0 citations · ▲ 337 on Hugging Face

Lower-Level Agnostic Bilevel Optimization

Peiwen Qiu, Prashant Khanduri, Jia (Kevin) Liu

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

Growing a Neural Network in Breadth, Depth, and Time

Eivinas Butkus, Kedar Garzón Gupta, Nikolaus Kriegeskorte

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

Sharpness-Aware Hybrid Model Learning for Architecture-Agnostic Parameter Estimation

Naoya Takeishi

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

Learn Locally, Recurse Globally: Neural Circuit Synthesis Beyond Training Depth

Emile Richard

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

Don't Waste Population: Post-Anneal Refinement for Combinatorial Optimization

Yuma Ichikawa

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

Richer or More Gates? Fan-In Trade-offs in Learnable Logic Circuits

Youngsung Kim

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

Conservative Pareto Set Amortization for Offline Multi-Objective Optimization

Ji Cheng

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

Learning Pareto Stationary Fronts via Single-Pass Backpropagation

Elina Rojin Celik, Marcos M. Raimundo, Isabel Valera

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

DASS: A Solver-Agnostic Dynamic Auxiliary Search Strategy for Symbolic Regression

Hu, Qian Li, Ding Wang, Yekun Zheng and 2 more

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

Frontier Task Synthesis Via Solution-Centric Evolution

Yangzhen Wu, Aaron Li, Wenjie Ma, Li Cao and 9 more

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

Scaling Arbitrary Architectures and Optimizers with Automatic Parameterization

Shikai Qiu, Charlie Chen, Andres Potapczynski, Martin Marek and 1 more

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

Treating Hyperparameters as Interventions: Task-Invariant Representation Learning for Transferable HPO

Mengyang Li, Ou Wu

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

A method to automatically discover symbolic local learning rules

Andrea Perin, Fabio Anselmi, Stephane Deny

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

Understanding Randomization in Greedy Model Search

Xin Chen, Jason Klusowski, Yan Shuo Tan, Chang Yu

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

Membrane Sensitivity and Deployment Fragility of Learnable Time Constants in Spiking Neural Networks

CHIU-CHANG CHENG, Ya-Ning Chang, Chao-Hung Wang

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

MOBO-CAPS: Multi-objective Bayesian Optimization with Cardinality-Aware Pareto Selection

Hanyang Wang, Juergen Branke, Matthias Poloczek

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

A Dynamic Decomposition Strategy for the MOEA/D With Proven Performance Guarantees

Benjamin Doerr, Martin S. Krejca, Noé Weeks

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

Strategic Feature Selection and Regularization

Jivat Neet Kaur, Pratik Patil, Divya Shanmugam, Emma Pierson and 5 more

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

Automated Causal Effect Estimation through Self-Evolving AI

Can Wang, Hongyu Zhao, Yiqun Chen

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

ML Configuration Artifacts Should Be Closed Under Review

Jan Rudy, Mohamed Khalil

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

NLD4CO: Neural Langevin Dynamics for Combinatorial Optimization

Jiale Ma, Wenzheng Pan, Binghao Cai, Xihe Zhang and 1 more

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

DeltaMomentum: A Key-Value based Anisotropic Momentum Update via Delta Rule

Euijin Hong, Guannan Qu

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

A Unified Merge Calculus for Learning-Rate Scaling on Neural Computation Graphs

Haosong Zhang, Wu Shenxi, Zhiyuan Che, Xi Chen and 2 more

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

ParetoM$^3$: Learning on the Pareto Set under Preference Guidance via Min-Max-Min Optimization

Pei Tang, Songtao Lu

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

Regularized Large Neighborhood Search

Regularized LNS turns local search heuristics into MCMC samplers with Fenchel-Young losses, enabling exact block Gibbs sampling and end-to-end learning without global solvers.

Germain Vivier-Ardisson, Laurent Demonet, Axel Parmentier, Mathieu Blondel

Paris Poster Session 6, Fri, Dec 11, 2:30 PM–4:30 PM, Paris Poster Hall · Published 2026

AMUSE: Anytime Muon with Stable Gradient Evaluation

AMUSE integrates Muon's rapid bulk progress with Schedule-Free averaging via time-varying interpolation to suppress oscillations, requiring no learning rate schedules and improving training efficiency across vision and LLM tasks.

Jueun Kim, Baekrok Shin, Jihun Yun, Beomhan Baek and 2 more

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

Layerwise LQR for Geometry-Aware Optimization of Deep Networks

Layerwise LQR frames deep network preconditioners as LQR problems to learn scalable structured inverse preconditioners preserving cross-layer geometry, improving optimization dynamics with modest overhead.

Simon Dufort-Labbé, Pierre-Luc Bacon, Razvan Pascanu, Simon Lacoste-Julien and 1 more

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

Automated Kernel Discovery Towards Understanding High-dimensional Bayesian Optimization

Kernel Discovery uses an LLM-driven evolutionary framework to search broad kernel spaces for high-dimensional Bayesian optimization, achieving average rank 1.2 out of 17.

Taeyoung Yun, Woocheol Shin, Inhyuck Song, Jaewoo Lee and 1 more

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

Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization

Reformulating constrained black-box optimization as latent-space posterior inference with flow-based surrogates and amortized diffusion sampling improves high-dimensional optimization performance.

Kiyoung Om, Kyuil Sim, Taeyoung Yun, Hyeongyu Kang and 1 more

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

Differentiable Knapsack and Top-k Operators via Dynamic Programming

A unified framework casts knapsack and top-k operators as dynamic programs with smoothed recursions for differentiable relaxations, parallel algorithms, and theoretical regularization guarantees.

Germain Vivier-Ardisson, Michael E Sander, Axel Parmentier, Mathieu Blondel

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

Quality-Diversity Optimization as Multi-Objective Optimization

Reformulating quality-diversity optimization as multi-objective optimization with many objectives enables set-based scalarization methods to solve QD problems with theoretical guarantees and competitive performance.

Xi Lin, Ping Guo, Yilu Liu, Bo Xue and 2 more

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

Solver-Aware Decompositions for Programming-by-Example: When Dividing Requires Knowing how to Conquer

Solver-Aware Decomposition trains PBE decomposers via synthesizer feedback, showing ground-truth subgoal alignment does not improve synthesis and optimizing for solver tractability yields consistent accuracy gains.

Janis Zenkner, Tobias Sesterhenn, Tim Grams, Christian Bartelt

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

Position: Let’s Strengthen Verifiability if We Can’t Enforce Reproducibility

Machine learning papers are hard to reproduce due to missing code, so researchers should prioritize verifiable results through concrete checkability improvements.

Samet Hicsonmez, Nermin Samet, Renaud Marlet

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

Position: Adopt Constraints Over Fixed Penalties in Deep Learning

Fixed weighted penalties poorly enforce hard deep learning constraints; use constrained formulations directly instead of scalarized surrogates.

Juan Ramirez, Seyed Meraj Hashemizadehaghda, Simon Lacoste-Julien

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

Learning to play with spikes. Characterizing, predicting, and engineering unsupervised plasticity rules for spiking reservoir computing

Local plasticity rules stabilize spiking reservoirs and structure representations into predictable, transferable signatures that allow direct engineering of high-performing neuromorphic algorithms.

Maciej Kania, Basile Confavreux, Tim Vogels

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

BROS: Bias-Corrected Randomized Subspaces for Memory-Efficient Single-Loop Bilevel Optimization

BROS proposes a memory-efficient single-loop bilevel optimization method using randomized subspaces and Rademacher bi-probe correction to recover unbiased Hessian estimates with O(ε^-2) sample complexity and up to 44.9% lower peak memory.

Hengrui Zhang, Boao Kong, Engao Zhang, Kun Yuan

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

Tree-Structured Synergy of Large Language Models and Bayesian Optimization for Efficient CASH

LB-MCTS combines tree-structured search with Bayesian optimization and large language models to solve CASH, outperforming baselines on 104 datasets via adaptive reliability-aware proposal shifts.

Beicheng Xu, Weitong Qian, Lingching Tung, Yupeng Lu and 1 more

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

MAdam: Metric-Aware Multi-Objective Adam

MAdam removes Adam's weighting and geometric mismatches in multi-objective optimization via a preference-conditioned curvature preconditioner, consistently improving results across tasks.

Fengbei Liu, Rachit Saluja, Sunwoo Kwak, Ruibo Wang and 4 more

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

An Open-Source Training Dataset for Foundation Models for Black-box Optimization

BBO-Pile provides 500K real-world black-box optimization trajectories across 3095 problems, and trained foundation models show large-scale pre-training effectively imitates optimization methods.

Aaron Klein, Herilalaina Rakotoarison, Luca Thale-Bombien, David Salinas

Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026 · ▲ 1 on Hugging Face

Zeroth-Order Sharpness-Aware Learning with Exponential Tilting

An exponential tilting objective unifies zeroth-order smoothing and sharpness-aware minimization, yielding gradient-free algorithms that improve generalization over baselines.

Xuchen Gong, Tian Li

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

Agentic Neural Architecture Search

AgentNAS uses LLMs to generate seed architectures decomposed into slotted scaffolds that define bounded search spaces for NAS, achieving state-of-the-art results on 11 of 17 diverse tasks.

Seokhoon Jeong, Mijung Kim, Taehwan Kim

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

Optimistic Dual Averaging Unifies Modern Optimizers

SODA unifies modern optimizers via optimistic dual averaging and improves them with a theoretical 1/k weight decay schedule requiring no tuning.

Thomas Pethick, Wanyun Xie, Roman Machacek, Volkan Cevher

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