4/23/2026
Spotlights and Blindspots: Evaluating Machine-Generated Text Detection
Evaluation of 15 machine-generated text detection models across 7 test sets to compare effectiveness and benchmark detection approaches.
4/23/2026
Mitigating Prompt-Induced Cognitive Biases in General-Purpose AI for Software Engineering
Research on prompt-induced cognitive biases in AI for software engineering decision support, showing how phrasing shifts affect model outputs.
4/23/2026
UCCL-Zip: Lossless Compression Supercharged GPU Communication
UCCL-Zip integrates lossless compression into GPU communication primitives for LLM training without numerical errors or convergence degradation.
4/23/2026
Atomic Decision Boundaries: A Structural Requirement for Guaranteeing Execution-Time Admissibility in Autonomous Systems
Research on atomic decision boundaries for enforcing admissibility constraints in autonomous systems at exact state transition moments.
4/23/2026
A multimodal and temporal foundation model for virtual patient representations at healthcare system scale
Apollo is multimodal temporal foundation model for patient representations trained on 30 years of longitudinal hospital data from 7.2M patients and 25B clinical records.
4/23/2026
Bounded Ratio Reinforcement Learning
Bounded Ratio Reinforcement Learning framework bridges theory-practice gap in PPO by formalizing trust region methods with bounded ratio constraints.
4/23/2026
LLM-as-Judge Framework for Evaluating Tone-Induced Hallucination in Vision-Language Models
Framework uses LLM-as-Judge to evaluate hallucinations in vision-language models under coercive prompt phrasing and linguistic pressure.
4/23/2026
Talking to a Know-It-All GPT or a Second-Guesser Claude? How Repair reveals unreliable Multi-Turn Behavior in LLMs
Study investigates how LLMs handle repair in multi-turn conversations, showing significant differences between models in self-initiated and user-initiated error correction.
4/23/2026
Cyber Defense Benchmark: Agentic Threat Hunting Evaluation for LLMs in SecOps
Cyber Defense Benchmark evaluates LLM agents on threat hunting tasks using 106 real attacks from MITRE ATT&CK, measuring SOC analyst capabilities on Windows event logs.
4/23/2026
RoLegalGEC: Legal Domain Grammatical Error Detection and Correction Dataset for Romanian
Dataset for grammatical error detection and correction in Romanian legal documents, trained on domain-specific legal text.
4/23/2026
WorkflowGen:an adaptive workflow generation mechanism driven by trajectory experience
WorkflowGen: adaptive workflow generation framework for LLM agents that reuses trajectory experience to reduce token consumption and improve execution stability.
4/23/2026
Transparent Screening for LLM Inference and Training Impacts
Framework for estimating environmental and computational impacts of LLM inference and training across market models with limited observability.
4/23/2026
Accelerating PayPal's Commerce Agent with Speculative Decoding: An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models
EAGLE3 speculative decoding optimization for PayPal's commerce agent, reducing latency on fine-tuned Nemotron-8B.
4/23/2026
On-Meter Graph Machine Learning: A Case Study of PV Power Forecasting for Grid Edge Intelligence
Graph neural networks for PV power forecasting on edge meters using GCN and GATv2 with ONNX deployment.
4/23/2026
Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts
Expert Upcycling: Mixture-of-Experts optimization shifting compute-efficient frontier for large language model scaling.
4/23/2026
Rethinking Reinforcement Fine-Tuning in LVLM: Convergence, Reward Decomposition, and Generalization
Theoretical analysis of reinforcement fine-tuning in vision-language models for agentic capabilities, convergence and generalization.
4/23/2026
DR-Venus: Towards Frontier Edge-Scale Deep Research Agents with Only 10K Open Data
DR-Venus: 4B open-data deep research agent for edge deployment, optimizing data quality and utilization for small models.
4/23/2026
Super Apriel: One Checkpoint, Many Speeds
Super Apriel: 15B-parameter supernet supporting multiple attention mechanisms switchable at inference time without reloading.
4/23/2026
Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
Sparse autoencoders reveal distinct LLM features encoding uncertainty versus correctness, showing functional dissociation.
4/23/2026
Multi-Objective Reinforcement Learning for Generating Covalent Inhibitor Candidates
Multi-objective reinforcement learning pipeline for rational design of covalent inhibitor candidates in drug discovery.
4/23/2026
Continuous Semantic Caching for Low-Cost LLM Serving
Semantic caching framework for reducing LLM inference costs by caching responses to semantically similar queries.
4/23/2026
Statistics, Not Scale: Modular Medical Dialogue with Bayesian Belief Engine
BMBE framework separating language and probabilistic reasoning in medical dialogue agents, using Bayesian belief engine for diagnostics.
4/23/2026
Federated Learning over Blockchain-Enabled Cloud Infrastructure
Integration of federated learning and blockchain technology for privacy-preserving machine learning in cloud infrastructure.
4/23/2026
Improved large-scale graph learning through ridge spectral sparsification
Graph Laplacian learning in distributed streaming settings using ridge spectral sparsification for large-scale graphs.
4/23/2026
On the Quantization Robustness of Diffusion Language Models in Coding Benchmarks
Investigation of post-training quantization robustness for diffusion-based language models on coding benchmarks with GPTQ and HAWQ.
4/23/2026
Energy-Based Open-Set Active Learning for Object Classification
Active learning framework for object classification handling open-set conditions with unknown classes using energy-based models.
4/23/2026
Differentiable Conformal Training for LLM Reasoning Factuality
Differentiable conformal training method to improve LLM factuality and reduce hallucinations with statistical guarantees on error rates.
4/23/2026
Learning to Solve the Quadratic Assignment Problem with Warm-Started MCMC Finetuning
Permutation learning framework (PLMA) combining neural networks with warm-started MCMC for solving quadratic assignment problems.
4/23/2026
Meta Additive Model: Interpretable Sparse Learning With Auto Weighting
Sparse additive model with automatic sample reweighting for high-dimensional analysis robust to non-Gaussian noise and outliers.
4/23/2026
On the Stability and Generalization of First-order Bilevel Minimax Optimization
Theoretical analysis of generalization and stability in first-order bilevel minimax optimization for hyperparameter optimization and reinforcement learning.
4/23/2026
Adaptive Conformal Anomaly Detection with Time Series Foundation Models for Signal Monitoring
Post-hoc adaptive conformal anomaly detection method leveraging pre-trained time series foundation models without fine-tuning for signal monitoring.
4/23/2026
A Delta-Aware Orchestration Framework for Scalable Multi-Agent Edge Computing
Framework addressing performance degradation when scaling multi-agent reinforcement learning beyond 100 agents in edge computing environments.
4/23/2026
Temporally Extended Mixture-of-Experts Models
Proposes temporally extended mixture-of-experts layers using reinforcement learning options framework to improve GPU memory efficiency during inference.
4/23/2026
SMART: A Spectral Transfer Approach to Multi-Task Learning
Spectral transfer method for multi-task linear regression that assumes spectral similarity between source and target models.
4/23/2026
Scaling Self-Play with Self-Guidance
Self-play scaling approach for LLMs with self-guidance to overcome learning plateaus and reward hacking.
4/23/2026
Machine Learning for Two-Stage Graph Sparsification for the Travelling Salesman Problem
ML methods for two-stage graph sparsification to improve TSP solver efficiency across instance sizes.
4/23/2026
Rethinking Intrinsic Dimension Estimation in Neural Representations
Analysis framework for intrinsic dimension estimation in neural network representations.
4/23/2026
Sheaf Neural Networks on SPD Manifolds: Second-Order Geometric Representation Learning
Sheaf neural networks on SPD manifolds for second-order geometric representation learning in graphs.
4/23/2026
Formalising the Logit Shift Induced by LoRA: A Technical Note
Formal analysis of logit shift effects from LoRA adaptation using first-order Fréchet approximation.
4/23/2026
R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling
R2IF: RL framework for LLM function calling with composite rewards aligning reasoning and tool-call decisions.
4/23/2026
Calibrating conditional risk
Methods for estimating expected loss of prediction models conditional on input features in classification and regression settings.
4/23/2026
Scalable AI Inference: Performance Analysis and Optimization of AI Model Serving
Performance analysis and optimization of BentoML-based AI inference system for scalable model serving in production environments.
4/23/2026
Efficient Test-Time Inference via Deterministic Exploration of Truncated Decoding Trees
Distinct Leaf Enumeration (DLE) deterministic decoding method efficiently explores truncated trees for test-time inference on math and code tasks.
4/23/2026
Explicit Dropout: Deterministic Regularization for Transformer Architectures
Deterministic dropout formulation expressing regularization as additive loss terms for Transformer architectures with explicit optimization objectives.
4/23/2026
CHASM: Unveiling Covert Advertisements on Chinese Social Media
CHASM dataset and benchmark for evaluating multimodal LLMs on detecting covert advertisements in Chinese social media moderation tasks.
4/23/2026
Differentially Private Clustered Federated Learning with Privacy-Preserving Initialization and Normality-Driven Aggregation
Federated learning framework with differential privacy and privacy-preserving initialization for heterogeneous cross-device deployments.
4/23/2026
Too Sharp, Too Sure: When Calibration Follows Curvature
Study of relationship between model calibration and loss surface curvature in neural networks, showing calibration emerges during training on vision tasks.
4/23/2026
Occupancy Reward Shaping: Improving Credit Assignment for Offline Goal-Conditioned Reinforcement Learning
Occupancy reward shaping method extracting temporal information from generative world models to improve credit assignment in offline goal-conditioned reinforcement learning.
4/23/2026
GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning
GRPO-VPS extends Group Relative Policy Optimization with verifiable process supervision to improve credit assignment for LLM reasoning tasks.
4/23/2026
Variance Is Not Importance: Structural Analysis of Transformer Compressibility Across Model Scales
Systematic empirical study of transformer compression across 40+ experiments on GPT-2 and Mistral 7B identifying five structural properties relevant to model compression.