4/21/2026
A Mechanism Study of Delayed Loss Spikes in Batch-Normalized Linear Models
Theoretical analysis of delayed loss spikes in batch-normalized linear models during neural network training.
4/21/2026
Towards Deep Encrypted Training: Low-Latency, Memory-Efficient, and High-Throughput Inference for Privacy-Preserving Neural Networks
Research on privacy-preserving neural network inference using homomorphic encryption for batch processing of encrypted inputs.
4/21/2026
Lorentz Framework for Semantic Segmentation
Lorentz hyperbolic space semantic segmentation framework with uncertainty quantification and numerical stability improvements.
4/21/2026
When Earth Foundation Models Meet Diffusion: An Application to Land Surface Temperature Super-Resolution
EFDiff framework uses Prithvi-EO-2.0 foundation model with diffusion for land surface temperature super-resolution.
4/21/2026
GRAIL: Autonomous Concept Grounding for Neuro-Symbolic Reinforcement Learning
GRAIL framework enables autonomous grounding of relational concepts in neuro-symbolic RL agents without manual definition.
4/21/2026
EasyVideoR1: Easier RL for Video Understanding
EasyVideoR1 extends reinforcement learning from verifiable rewards to video understanding in multimodal models.
4/21/2026
Freshness-Aware Prioritized Experience Replay for LLM/VLM Reinforcement Learning
Freshness-aware prioritized experience replay improves sample efficiency in off-policy RL for LLM/VLM post-training.
4/21/2026
LLMs can persuade only psychologically susceptible humans on societal issues, via trust in AI and emotional appeals, amid logical fallacies
Talk2AI longitudinal study measuring LLM persuasiveness on societal topics via psychological and reasoning dimensions.
4/21/2026
Bolzano: Case Studies in LLM-Assisted Mathematical Research
Bolzano multi-agent LLM system produces novel mathematics results through parallel provers and verifier with persistent knowledge base.
4/21/2026
SPS: Steering Probability Squeezing for Better Exploration in Reinforcement Learning for Large Language Models
SPS method improves exploration in RL for LLMs by addressing probability squeezing limitations in reasoning model training.
4/21/2026
Improving LLM Code Reasoning via Semantic Equivalence Self-Play with Formal Verification
Self-play framework for improving LLM code reasoning via semantic equivalence validation using formal verification and Liquid Haskell proofs.
4/21/2026
Harness as an Asset: Enforcing Determinism via the Convergent AI Agent Framework (CAAF)
CAAF framework enforces determinism in LLM-based agentic workflows for safety-critical systems using convergent execution patterns.
4/21/2026
Trajectory-Restricted Optimization Conditions and Geometry-Aware Linear Convergence
Develops trajectory-restricted framework for analyzing linear convergence of first-order optimization methods with geometry-aware bounds.
4/21/2026
Stability-Weighted Decoding for Diffusion Language Models
Proposes stability-weighted decoding for diffusion language models using temporal token stability to improve parallel text generation.
4/21/2026
EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling
EvoComp reduces visual token count in multimodal LLMs while preserving accuracy using semantic-guided evolutionary compression.
4/21/2026
Automated Classification of Plasma Regions at Mars Using Machine Learning
Develops ML-based classifier to automatically identify plasma regions around Mars (solar wind, magnetosheath, magnetosphere) from mission data.
4/21/2026
Local Inconsistency Resolution: The Interplay between Attention and Control in Probabilistic Models
Presents Local Inconsistency Resolution algorithm for learning and inference in probabilistic models using Probabilistic Dependency Graphs.
4/21/2026
FlowRefiner: Flow Matching-Based Iterative Refinement for 3D Turbulent Flow Simulation
FlowRefiner uses flow matching for iterative refinement of 3D turbulent flow neural PDE solvers with improved autoregressive prediction.
4/21/2026
SynthFix: Adaptive Neuro-Symbolic Code Vulnerability Repair
SynthFix combines LLMs with symbolic AI and compiler feedback for automated code vulnerability repair using neural-symbolic hybrid approach.
4/21/2026
Bilinear Input Modulation for Mamba: Koopman Bilinear Forms for Memory Retention and Multiplicative Computation
Proposes bilinear input modulation for Mamba SSMs to improve memory retention and computational capacity using Koopman bilinear forms.
4/21/2026
Bit-Flip Vulnerability of Shared KV-Cache Blocks in LLM Serving Systems
Security analysis of bit-flip vulnerabilities in shared KV-cache blocks within vLLM's prefix caching for LLM serving systems.
4/21/2026
RoTRAG: Rule of Thumb Reasoning for Conversation Harm Detection with Retrieval-Augmented Generation
RoTRAG: retrieval-augmented generation method for multi-turn dialogue harm detection using explicit normative principles and rules of thumb.
4/21/2026
ARMove: Learning to Predict Human Mobility through Agentic Reasoning
ARMove: agentic reasoning framework for human mobility prediction using LLMs with improved interpretability, iterative learning, and transferability.
4/21/2026
Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning
Hybrid ensemble method combining Chain-of-Thought and Program-of-Thought reasoning to achieve self-consistency with only two LLM samples.
4/21/2026
Waking Up Blind: Cold-Start Optimization of Supervision-Free Agentic Trajectories for Grounded Visual Perception
SPECTRA: supervision-free framework using cold-start reinforcement learning to improve vision-language model agents' visual perception and tool use.
4/21/2026
Matlas: A Semantic Search Engine for Mathematics
Semantic search engine for retrieving mathematical knowledge across millions of documents to ground AI mathematics systems.
4/21/2026
ONTO: A Token-Efficient Columnar Notation for LLM Input Optimization
ONTO: token-efficient columnar notation for serializing operational data to LLMs, reducing JSON overhead by ~80% in IoT sensor datasets.
4/21/2026
Target Parameterization in Diffusion Models for Nonlinear Spatiotemporal System Identification
Diffusion model approach for identifying parameters in nonlinear spatiotemporal systems with improved robustness in turbulent regimes.
4/21/2026
Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity
Study showing LLM-based agents fail to incorporate unexpected environmental observations into reasoning across three benchmarks, limiting their adaptability.
4/21/2026
Characterizing Model-Native Skills
Research on characterizing language model skills through model-native representations rather than external taxonomies for intervention on model behavior.
4/21/2026
Improving reproducibility by controlling random seed stability in machine learning based estimation via bagging
Theoretical and practical methods for improving reproducibility in ML estimation by controlling random seed instability via bagging approaches.
4/21/2026
Program Structure-aware Language Models: Targeted Software Testing beyond Textual Semantics
GLMTest framework for LLM-based test generation using program structure awareness to target high-risk branches and improve bug discovery.
4/21/2026
Concurrent Criterion Validation of a Validity Screen for LLM Confidence Signals via Selective Prediction
Validity screen framework for LLM confidence signals evaluated on 20 frontier models, showing predictive validity for selective prediction performance.
4/21/2026
A Quasi-Experimental Developer Study of Security Training in LLM-Assisted Web Application Development
Controlled developer study measuring security training effectiveness on LLM-assisted Java Spring Boot implementation with within and between-subject design.
4/21/2026
Forget What Matters, Keep the Rest: Selective Unlearning of Informative Tokens
Token-level selective unlearning for LLMs using importance-weighted forgetting to maintain model utility while removing harmful information.
4/21/2026
Adversarial Arena: Crowdsourcing Data Generation through Interactive Competition
Dataset generation method using adversarial competition between attackers and defenders to create diverse, high-quality conversational training data.
4/21/2026
Scaling Human-AI Coding Collaboration Requires a Governable Consensus Layer
Analysis of AI-assisted code generation failures as control problems, proposing governance layers for tracking structural commitments in human-AI collaboration.
4/21/2026
Physics-Informed Causal MDPs for Sequential Constraint Repair in Engineering Simulation Pipelines
Framework for off-policy learning in constrained MDPs with layered dependency structures for engineering simulation repair tasks.
4/21/2026
Prompting Foundation Models for Zero-Shot Ship Instance Segmentation in SAR Imagery
Zero-shot ship detection in SAR imagery using foundation models and YOLO, demonstrating prompting approaches for specialized domain tasks.
4/21/2026
Heterogeneity in Formal Linguistic Competence of Language Models: Is Data the Real Bottleneck?
Study examining whether LLM failures in formal linguistic competence stem from architecture limits or data scarcity via controlled pretraining experiments.
4/21/2026
A Sugeno Integral View of Binarized Neural Network Inference
Mathematical connection between binarized neural networks and Sugeno integrals, showing BNN inference as rule-based systems.
4/21/2026
How Creative Are Large Language Models in Generating Molecules?
Research on using LLMs for molecular generation under chemical constraints, framing creativity as a functional requirement for exploring large chemical spaces.
4/21/2026
Depth Registers Unlock W4A4 on SwiGLU: A Reader/Generator Decomposition
Depth Registers method for W4A4 post-training quantization in language models, reducing error from 1727x to 14x on SwiGLU models.
4/21/2026
Soft Label Pruning and Quantization for Large-Scale Dataset Distillation
Dataset distillation technique using soft label pruning and quantization to reduce storage overhead on large-scale ImageNet datasets.
4/21/2026
AQPIM: Breaking the PIM Capacity Wall for LLMs with In-Memory Activation Quantization
Activation quantization technique (AQPIM) for running large LLMs on Processing-in-Memory architectures with reduced memory footprint.
4/21/2026
Distributional Off-Policy Evaluation with Deep Quantile Process Regression
Distributional off-policy evaluation using deep quantile process regression for estimating return distributions in reinforcement learning.
4/21/2026
mlr3torch: A Deep Learning Framework in R based on mlr3 and torch
R package mlr3torch for deep learning on tabular data and tensors, integrating torch with mlr3 ecosystem for classification and regression.
4/21/2026
DiffuSAM: Diffusion Guided Zero-Shot Object Grounding for Remote Sensing Imagery
Diffusion model pipeline for zero-shot object grounding in remote sensing imagery, combining diffusion with SAM segmentation models.
4/21/2026
Domain-Specialized Object Detection via Model-Level Mixtures of Experts
Mixture-of-Experts applied to object detection for domain-specialized models with improved interpretability over standard ensembles.
4/21/2026
Incremental learning for audio classification with Hebbian Deep Neural Networks
Hebbian learning approach for continual audio classification using kernel plasticity to balance learning new information with retaining previous knowledge.