3/20/2026
Engineering Verifiable Modularity in Transformers via Per-Layer Supervision
Architectural approach using per-layer supervision to expose hidden modularity in Transformers, enabling interpretability and causal control of components.
3/20/2026
Towards Differentiating Between Failures and Domain Shifts in Industrial Data Streams
Methods for distinguishing system failures from domain shifts in industrial data streams using anomaly detection techniques.
3/20/2026
Adapting Methods for Domain-Specific Japanese Small LMs: Scale, Architecture, and Quantization
Systematic methodology for fine-tuning domain-specific Japanese small language models, identifying optimal training scale (4k samples), base models, and quantization strategies.
3/20/2026
NANOZK: Layerwise Zero-Knowledge Proofs for Verifiable Large Language Model Inference
Zero-knowledge proof system enabling cryptographic verification that proprietary LLM API outputs come from claimed models, preventing model substitution or degradation.
3/20/2026
Fundamental Limits of Neural Network Sparsification: Evidence from Catastrophic Interpretability Collapse
Investigates how mechanistic interpretability features survive extreme neural network sparsification using adaptive sparsity scheduling in VAE-SAE architectures.
3/20/2026
Lightweight Adaptation for LLM-based Technical Service Agent: Latent Logic Augmentation and Robust Noise Reduction
Lightweight adaptation method for LLM-based technical service agents using latent logic augmentation and noise reduction without full retraining.
3/20/2026
SLEA-RL: Step-Level Experience Augmented Reinforcement Learning for Multi-Turn Agentic Training
Step-level experience augmented reinforcement learning for multi-turn LLM agents that dynamically retrieve and refine experiences throughout episodes.
3/20/2026
Probabilistic Federated Learning on Uncertain and Heterogeneous Data with Model Personalization
Meta-BayFL framework for federated learning with probabilistic approaches to handle data uncertainty and heterogeneity while managing computational overhead.
3/20/2026
Enhancing Reinforcement Learning Fine-Tuning with an Online Refiner
Proposes dynamic constraints for reinforcement learning fine-tuning that adapt to model capabilities, resolving tension between optimization and constraint satisfaction.
3/20/2026
ARTEMIS: A Neuro Symbolic Framework for Economically Constrained Market Dynamics
Neuro-symbolic framework combining neural operators with economic constraints for interpretable quantitative finance models respecting no-arbitrage principles.
3/20/2026
Tula: Optimizing Time, Cost, and Generalization in Distributed Large-Batch Training
Tula optimizes distributed large-batch training by balancing communication overhead, computation cost, and model generalization across scaling configurations.
3/20/2026
VC-Soup: Value-Consistency Guided Multi-Value Alignment for Large Language Models
Proposes VC-Soup method for aligning LLMs with multiple potentially conflicting human values through value-consistency guided optimization.
3/20/2026
LLM-Augmented Computational Phenotyping of Long Covid
LLM-augmented computational phenotyping framework for discovering clinical subphenotypes in Long COVID through iterative hypothesis generation and evidence extraction.
3/20/2026
Conflict-Free Policy Languages for Probabilistic ML Predicates: A Framework and Case Study with the Semantic Router DSL
Framework for detecting conflicts in policy languages that use probabilistic ML predicates, applied to semantic router DSL for LLM routing systems.
3/20/2026
Gradient-Informed Temporal Sampling Improves Rollout Accuracy in PDE Surrogate Training
Improves PDE surrogate model training through gradient-informed temporal sampling strategies that optimize rollout accuracy under fixed data budgets.
3/20/2026
MolRGen: A Training and Evaluation Setting for De Novo Molecular Generation with Reasonning Models
Framework for evaluating reasoning-based LLMs on de novo molecular generation and drug discovery without requiring ground-truth molecule pairs.
3/20/2026
Discovering What You Can Control: Interventional Boundary Discovery for Reinforcement Learning
Proposes Interventional Boundary Discovery to identify causal state dimensions agents can control, using Pearl's do-operator for causal identification.
3/20/2026
Sharpness-Aware Minimization in Logit Space Efficiently Enhances Direct Preference Optimization
Addresses the squeezing effect in Direct Preference Optimization (DPO) for LLM alignment using sharpness-aware minimization in logit space.
3/20/2026
Detection Is Cheap, Routing Is Learned: Why Refusal-Based Alignment Evaluation Fails
Studies alignment evaluation in LLMs by examining political censorship in Chinese language models, focusing on routing mechanisms beyond concept detection and refusal behaviors.
3/20/2026
On Additive Gaussian Processes for Wind Farm Power Prediction
Additive Gaussian processes for wind farm power prediction using population-based structural health monitoring perspective.
3/20/2026
Path-Constrained Mixture-of-Experts
Path-constrained mixture-of-experts architecture constraining expert routing paths to improve statistical efficiency and meaningful parameter structure.
3/20/2026
ALIGN: Adversarial Learning for Generalizable Speech Neuroprosthesis
ALIGN: adversarial learning framework for session-invariant speech neuroprosthesis decoding from brain-computer interfaces.
3/20/2026
Approximate Subgraph Matching with Neural Graph Representations and Reinforcement Learning
Neural graph representation learning with RL for approximate subgraph matching, an NP-hard problem in graph analysis.
3/20/2026
Learning to Reason with Curriculum I: Provable Benefits of Autocurriculum
Autocurriculum training methods with provable benefits for chain-of-thought reasoning in language models with reduced data/compute costs.
3/20/2026
Escaping Offline Pessimism: Vector-Field Reward Shaping for Safe Frontier Exploration
Vector-field reward shaping for offline RL to enable safe exploration near dataset boundaries using simulator confidence.
3/20/2026
Epistemic Generative Adversarial Networks
Epistemic GANs using Dempster-Shafer theory to improve output diversity and architectural enhancements for generative models.
3/20/2026
Mathematical Foundations of Deep Learning
Comprehensive book on mathematical foundations of deep learning covering neural network approximation theory, optimal control, RL, and generative models.
3/20/2026
RE-SAC: Disentangling aleatoric and epistemic risks in bus fleet control: A stable and robust ensemble DRL approach
RE-SAC: ensemble deep reinforcement learning for bus fleet control that disentangles aleatoric and epistemic uncertainty.
3/20/2026
FlowMS: Flow Matching for De Novo Structure Elucidation from Mass Spectra
Flow matching approach for de novo molecular structure elucidation from mass spectra using deep generative models.
3/20/2026
Self-Tuning Sparse Attention: Multi-Fidelity Hyperparameter Optimization for Transformer Acceleration
AFBS-BO framework for automated hyperparameter optimization of sparse attention mechanisms in transformers via adaptive fidelity Bayesian optimization.
3/20/2026
Towards Noise-Resilient Quantum Multi-Armed and Stochastic Linear Bandits
Quantum multi-armed and stochastic linear bandits algorithms robust to noise in NISQ devices, achieving quadratic speedups over classical methods.
3/20/2026
Discounted Beta--Bernoulli Reward Estimation for Sample-Efficient Reinforcement Learning with Verifiable Rewards
Sample-efficient reward estimation method for RL with verifiable rewards in large language model post-training.
3/20/2026
Seeking Universal Shot Language Understanding Solutions
Training suite for film shot language understanding using vision-language models to match expert cinematographic analysis.
3/20/2026
AcceRL: A Distributed Asynchronous Reinforcement Learning and World Model Framework for Vision-Language-Action Models
Distributed asynchronous RL framework for Vision-Language-Action models with integrated trainable world models.
3/20/2026
AIMER: Calibration-Free Task-Agnostic MoE Pruning
Calibration-free pruning method for Mixture-of-Experts language models to reduce memory and serving overhead.
3/20/2026
Balancing the Reasoning Load: Difficulty-Differentiated Policy Optimization with Length Redistribution for Efficient and Robust Reinforcement Learning
Policy optimization approach addressing overthinking in large reasoning models through difficulty-differentiated training.
3/20/2026
Data-efficient pre-training by scaling synthetic megadocs
Study on synthetic data augmentation for efficient pre-training with better loss scaling using synthetic megadocs.
3/20/2026
Beyond Passive Aggregation: Active Auditing and Topology-Aware Defense in Decentralized Federated Learning
Research on active auditing framework against backdoor attacks in decentralized federated learning systems.
3/20/2026
GAPSL: A Gradient-Aligned Parallel Split Learning on Heterogeneous Data
GAPSL: gradient-aligned parallel split learning for federated learning on heterogeneous data, reducing client computational load.
3/20/2026
SINDy-KANs: Sparse identification of non-linear dynamics through Kolmogorov-Arnold networks
SINDy-KANs combines Kolmogorov-Arnold networks with sparse identification to learn interpretable equations for nonlinear dynamical systems.
3/20/2026
Transformers Learn Robust In-Context Regression under Distributional Uncertainty
Shows Transformers learn robust in-context regression under distributional uncertainty without restrictive assumptions on data and noise.
3/20/2026
SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decoding
SpecForge: open-source training framework for speculative decoding draft models, improving LLM inference latency through token batching.
3/20/2026
Attack by Unlearning: Unlearning-Induced Adversarial Attacks on Graph Neural Networks
Demonstrates adversarial attacks on GNNs exploitable through unlearning mechanisms designed for GDPR compliance in graph learning systems.
3/20/2026
Elastic Weight Consolidation Done Right for Continual Learning
Systematic analysis of Elastic Weight Consolidation for continual learning, identifying issues with importance estimation and weight regularization methods.
3/20/2026
Evaluating Model-Free Policy Optimization in Masked-Action Environments via an Exact Blackjack Oracle
Evaluates model-free policy optimization algorithms using exact blackjack oracle with ground-truth benchmarks for discrete stochastic control.
3/20/2026
Revisiting Label Inference Attacks in Vertical Federated Learning: Why They Are Vulnerable and How to Defend
Studies label inference attacks in vertical federated learning, analyzing vulnerabilities when passive parties infer active party's labels and proposing defenses.
3/20/2026
HISR: Hindsight Information Modulated Segmental Process Rewards For Multi-turn Agentic Reinforcement Learning
HISR proposes segmental process rewards for multi-turn RL in LLM agents, addressing sparse reward propagation and credit assignment in long-horizon decision-making tasks.
3/20/2026
STEP: Scientific Time-Series Encoder Pretraining via Cross-Domain Distillation
Investigates transfer learning from audio and time-series foundation models to scientific time-series via cross-domain distillation.
3/20/2026
CausalRM: Causal-Theoretic Reward Modeling for RLHF from Observational User Feedbacks
Causal-theoretic approach for reward modeling using observational user feedback instead of expensive annotated data for RLHF alignment.
3/20/2026
Are complicated loss functions necessary for teaching LLMs to reason?
Ablation study examining necessity of components in Group Relative Policy Optimization for teaching LLMs reasoning and mathematical ability.