4/1/2026
Expressive Power of Implicit Models: Rich Equilibria and Test-Time Scaling
Analysis of implicit models with infinite-depth weight-tied networks that match explicit models while reducing memory consumption.
4/1/2026
When Does Global Attention Help? A Unified Empirical Study on Atomistic Graph Learning
Empirical study comparing message passing neural networks and graph transformers for atomistic property prediction.
4/1/2026
The Effect of Attention Head Count on Transformer Approximation
Theoretical analysis of how attention head count influences transformer approximation properties and expressive power.
4/1/2026
ARROW: An Adaptive Rollout and Routing Method for Global Weather Forecasting
Adaptive rollout and routing method for data-driven weather forecasting with improved spatiotemporal modeling.
4/1/2026
A Semi-amortized Lifted Learning-to-Optimize Masked (SALLO-M) Transformer Model for Scalable and Generalizable Beamforming
Learning-to-optimize Transformer framework for scalable beamforming in multi-user wireless systems.
4/1/2026
Automated Algorithm Design for Auto-Tuning Optimizers
Automated algorithm design using machine learning to optimize hyperparameter auto-tuning for high-performance applications.
4/1/2026
DeepCoT: Deep Continual Transformers for Real-Time Inference on Data Streams
Continual Transformers architecture for real-time low-latency inference on streaming data with reduced redundant computation.
4/1/2026
Deep Unfolding: Recent Developments, Theory, and Design Guidelines
Survey of deep unfolding techniques combining classical optimization algorithms with neural networks for signal processing.
4/1/2026
Stronger Normalization-Free Transformers
Research on normalization-free transformer architectures using Dynamic Tanh as alternative to standard normalization layers.
4/1/2026
Provably Extracting the Features from a General Superposition
Learning-theoretic approach to extracting interpretable features from superposition in complex ML models.
4/1/2026
HOSL: Hybrid-Order Split Learning for Memory-Constrained Edge Training
Split learning system using hybrid-order optimization to reduce memory overhead for collaborative LLM training on edge devices.
4/1/2026
The Mouth is Not the Brain: Bridging Energy-Based World Models and Language Generation
Architecture separating energy-based world models from language generation in LLMs to improve understanding vs. fluency tradeoff.
4/1/2026
Sparsity-Aware Unlearning for Large Language Models
Addresses machine unlearning for sparse LLMs to remove memorized sensitive information while maintaining model sparsification benefits for efficient deployment.
4/1/2026
ECHO-2: A Large-Scale Distributed Rollout Framework for Cost-Efficient Reinforcement Learning
ECHO-2 is distributed RL framework for LLM post-training via reinforcement learning, optimizing cost-efficiency of rollout generation across distributed resources.
4/1/2026
Joint Embedding Variational Bayes
VJE introduces reconstruction-free latent-variable framework for self-supervised learning using symmetric conditional ELBO on paired embeddings.
4/1/2026
A Fast and Generalizable Fourier Neural Operator-Based Surrogate for Melt-Pool Prediction in Laser Processing
LP-FNO uses Fourier Neural Operators as surrogate model for laser welding simulations, enabling faster parametric solution learning for industrial process optimization.
4/1/2026
DGPO: RL-Steered Graph Diffusion for Neural Architecture Generation
DGPO: RL-guided graph diffusion model for neural architecture search using reinforcement learning steering.
4/1/2026
Not All News Is Equal: Topic- and Event-Conditional Sentiment from Finetuned LLMs for Aluminum Price Forecasting
Study using finetuned LLMs for topic-conditional sentiment extraction to forecast aluminum commodity prices.
4/1/2026
Privacy-Preserving Machine Learning for IoT: A Cross-Paradigm Survey and Future Roadmap
Survey of privacy-preserving ML techniques for IoT including federated learning and differential privacy approaches.
4/1/2026
Sample-Efficient Hypergradient Estimation for Decentralized Bi-Level Reinforcement Learning
Decentralized bi-level RL algorithm for environment design with sample-efficient hypergradient estimation.
4/1/2026
MDM-Prime-v2: Binary Encoding and Index Shuffling Enable Compute-optimal Scaling of Diffusion Language Models
MDM-Prime-v2: Improvements to masked diffusion language models through binary encoding and index shuffling.
4/1/2026
FIPO: Eliciting Deep Reasoning with Future-KL Influenced Policy Optimization
FIPO: RL algorithm improving token-level credit assignment for reasoning in LLMs beyond outcome-based rewards.
4/1/2026
Beyond Hard Constraints: Budget-Conditioned Reachability For Safe Offline Reinforcement Learning
Safety-aware offline RL method using budget-conditioned reachability analysis for constrained decision-making.
4/1/2026
SkillRouter: Skill Routing for LLM Agents at Scale
SkillRouter: System for routing LLM agent requests to relevant skills from large skill libraries at inference time.
4/1/2026
ITQ3_S: High-Fidelity 3-bit LLM Inference via Interleaved Ternary Quantization with Rotation-Domain Smoothing
ITQ3_S: 3-bit LLM quantization method using interleaved ternary quantization and rotation-domain smoothing for efficient inference.
4/1/2026
Principal Prototype Analysis on Manifold for Interpretable Reinforcement Learning
Interpretability method for reinforcement learning using principal prototype analysis on manifolds.
4/1/2026
Image-Specific Adaptation of Transformer Encoders for Compute-Efficient Segmentation
Vision transformer optimization for image segmentation with adaptive computation per input image.
4/1/2026
EventChat: Implementation and user-centric evaluation of a large language model-driven conversational recommender system for exploring leisure events in an SME context
LLM-driven conversational recommender system for leisure event discovery with user-centric evaluation in SME context.
4/1/2026
Image Segmentation via Divisive Normalization: dealing with environmental diversity
Image segmentation approach using divisive normalization for autonomous driving under diverse environmental conditions.
4/1/2026
Tightening convex relaxations of trained neural networks: a unified approach for convex and S-shaped activations
Framework for tightening convex relaxations of trained neural networks with convex and S-shaped activations for optimization incorporation.
4/1/2026
Real-Time Operator Takeover for Visuomotor Diffusion Policy Training
Real-time operator takeover paradigm allowing seamless human intervention and correction during visuomotor diffusion policy execution.
4/1/2026
Aleph-Alpha-GermanWeb: Improving German-language LLM pre-training with model-based data curation and synthetic data generation
German-language LLM pre-training dataset curated via heuristic filtering, model-based selection, and synthetic data generation.
4/1/2026
LLM-Meta-SR: In-Context Learning for Evolving Selection Operators in Symbolic Regression
Meta-learning framework using LLMs to automatically design selection operators for evolutionary symbolic regression algorithms.
4/1/2026
AVA-Bench: Atomic Visual Ability Benchmark for Vision Foundation Models
AVA-Bench systematically evaluates atomic visual abilities of vision foundation models independent of LLM instruction tuning.
4/1/2026
Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles
SlowFast Sampling optimizes inference efficiency in diffusion-based language models through dynamic, flexible token generation strategies.
4/1/2026
Streaming 4D Visual Geometry Transformer
Streaming transformer architecture inspired by autoregressive LLMs for real-time 3D geometry perception and reconstruction from video.
4/1/2026
NES: An Instruction-Free, Low-Latency Next Edit Suggestion Framework Powered by Learned Historical Editing Trajectories
NES is an instruction-free code editing framework that learns from historical editing trajectories to suggest next edits with low latency.
4/1/2026
TTA-DAME: Test-Time Adaptation with Domain Augmentation and Model Ensemble for Dynamic Driving Conditions
Test-time adaptation method using domain augmentation and model ensembles to handle weather-related domain shifts in autonomous driving.
4/1/2026
Multi-Level Knowledge Distillation and Dynamic Self-Supervised Learning for Continual Learning
Knowledge distillation and self-supervised learning approach for continual learning with class-incremental learning and external unlabeled data.
4/1/2026
A Machine Learning Based Explainability Framework for Interpreting Swarm Intelligence
Interpretability framework for understanding how components of particle swarm optimization algorithms affect performance.
4/1/2026
ORIC: Benchmarking Object Recognition under Contextual Incongruity in Large Vision-Language Models
Benchmark evaluating how large vision-language models handle object recognition in contextually incongruent scenes and manage uncertainty.
4/1/2026
ProxyAttn: Guided Sparse Attention via Representative Heads
ProxyAttn method using representative attention heads to enable efficient sparse attention in LLMs for long-text processing with minimal performance degradation.
4/1/2026
MSG: Multi-Stream Generative Policies for Sample-Efficient Robotic Manipulation
Multi-Stream Generative Policy framework for robot learning that combines multiple object-centric policies at inference to improve sample efficiency and generalization.
4/1/2026
Symbol Grounding in Neuro-Symbolic AI: A Gentle Introduction to Reasoning Shortcuts
Neuro-symbolic AI overview connecting neural networks with symbolic reasoning to satisfy constraints, addressing reliable trustworthy AI development.
4/1/2026
Local Causal Discovery for Statistically Efficient Causal Inference
Efficient local causal discovery method for identifying adjustment sets without learning the full causal graph.
4/1/2026
VLA Models Are More Generalizable Than You Think: Revisiting Physical and Spatial Modeling
One-shot adaptation framework improving vision-language-action model generalization to novel camera viewpoints through spatial representation recalibration.
4/1/2026
Script Gap: Evaluating LLM Triage on Indian Languages in Native vs Romanized Scripts in a Real World Setting
Evaluation of LLM performance on Indian language maternal healthcare triage, comparing native scripts versus romanized text in real-world deployment.
4/1/2026
Guiding a Diffusion Transformer with the Internal Dynamics of Itself
Guidance strategy for diffusion transformers using internal model dynamics to improve image generation quality without external classifiers.
4/1/2026
The Geometry of Thought: How Scale Restructures Reasoning In Large Language Models
Analysis of 25k chain-of-thought trajectories showing neural scaling triggers domain-specific phase transitions in reasoning rather than uniform capability improvements across 8B-70B parameter models.
4/1/2026
$V_0$: A Generalist Value Model for Any Policy at State Zero
V0 is a generalist value model for policy gradient methods that scales efficiently with LLM training, replacing large critic models in actor-critic methods like PPO.