3/6/2026
A Benchmark Study of Neural Network Compression Methods for Hyperspectral Image Classification
Benchmark study comparing neural network compression techniques for hyperspectral image classification on edge devices.
3/6/2026
Model Medicine: A Clinical Framework for Understanding, Diagnosing, and Treating AI Models
Model Medicine: research framework for understanding, diagnosing, and treating AI model disorders through biological analogies.
3/6/2026
Interactive Benchmarks
Interactive Benchmarks: evaluation framework assessing model reasoning ability through active information acquisition under constraints.
3/6/2026
CONE: Embeddings for Complex Numerical Data Preserving Unit and Variable Semantics
CONE: hybrid transformer embedding method for preserving numerical data semantics and units in large language models.
3/6/2026
Evaluating GPT-5 as a Multimodal Clinical Reasoner: A Landscape Commentary
Evaluation of GPT-5 family models on multimodal clinical reasoning tasks including diagnosis with patient data and imaging.
3/6/2026
Guiding Diffusion-based Reconstruction with Contrastive Signals for Balanced Visual Representation
Method using diffusion models with contrastive signals to improve visual representations from CLIP encoders.
3/6/2026
The Inductive Bias of Convolutional Neural Networks: Locality and Weight Sharing Reshape Implicit Regularization
Machine learning research on how convolutional neural network architecture affects implicit regularization and generalization.
3/6/2026
WhisperAlign: Word-Boundary-Aware ASR and WhisperX-Anchored Pyannote Diarization for Long-Form Bengali Speech
Speech recognition and speaker diarization system for Bengali long-form audio using Whisper and Pyannote models.
3/6/2026
On the Strengths and Weaknesses of Data for Open-set Embodied Assistance
Study on data requirements and generalization of embodied foundation models for robotics and autonomous driving in interactive settings.
3/6/2026
Mitigating Instance Entanglement in Instance-Dependent Partial Label Learning
Machine learning research on partial label learning with instance-dependent label ambiguity, addressing instance entanglement challenges.
3/6/2026
Osmosis Distillation: Model Hijacking with the Fewest Samples
Research on security threats in transfer learning combined with dataset distillation, proposing osmosis distillation for model hijacking with minimal samples.
3/6/2026
Causally Robust Reward Learning from Reason-Augmented Preference Feedback
ReCouPLe framework uses natural language to mitigate causal confusion in preference-based reward learning for AI agents, improving robustness against spurious correlations.
3/6/2026
How Does the ReLU Activation Affect the Implicit Bias of Gradient Descent on High-dimensional Neural Network Regression?
Characterizes implicit bias of gradient descent on high-dimensional ReLU neural network regression with random features.
3/6/2026
VPWEM: Non-Markovian Visuomotor Policy with Working and Episodic Memory
Visuomotor policy with working and episodic memory for non-Markovian robotic control tasks from imitation learning.
3/6/2026
TimeWarp: Evaluating Web Agents by Revisiting the Past
TimeWarp benchmark evaluates web agents across six UI versions spanning different internet eras using containerized environments.
3/6/2026
Replaying pre-training data improves fine-tuning
Shows that replaying pre-training data during fine-tuning improves LLM performance on target domains while preventing catastrophic forgetting.
3/6/2026
Measuring the Fragility of Trust: Devising Credibility Index via Explanation Stability (CIES) for Business Decision Support Systems
CIES metric measures robustness and stability of XAI explanations under data perturbations for business decision systems.
3/6/2026
RepoLaunch: Automating Build&Test Pipeline of Code Repositories on ANY Language and ANY Platform
RepoLaunch agent automates dependency resolution, compilation, and test extraction for software repositories across languages and platforms.
3/6/2026
Good-Enough LLM Obfuscation (GELO)
GELO obfuscation scheme protects LLM prompt privacy on shared accelerators against memory access attacks.
3/6/2026
ARC-TGI: Human-Validated Task Generators with Reasoning Chain Templates for ARC-AGI
ARC-TGI open-source framework for task-family generators that create diverse ARC-AGI puzzles with reasoning chain templates.
3/6/2026
SRasP: Self-Reorientation Adversarial Style Perturbation for Cross-Domain Few-Shot Learning
Novel style perturbation method for cross-domain few-shot learning addressing gradient instability and sharp minima convergence.
3/6/2026
Feature Resemblance: On the Theoretical Understanding of Analogical Reasoning in Transformers
Theoretical analysis of analogical reasoning emergence in transformers through aligned representations and sequential training.
3/6/2026
Balancing Coverage and Draft Latency in Vocabulary Trimming for Faster Speculative Decoding
Optimizes vocabulary trimming in draft models for speculative decoding to balance inference latency and token coverage in LLM acceleration.
3/6/2026
KARL: Knowledge Agents via Reinforcement Learning
KARL system trains enterprise search agents via reinforcement learning with KARLBench evaluation suite covering six search tasks including entity search and report synthesis.
3/6/2026
Learning Optimal Individualized Decision Rules with Conditional Demographic Parity
Proposes methods for training individualized decision rules with fairness constraints to mitigate discriminatory bias in personalized systems.
3/6/2026
Boosting ASR Robustness via Test-Time Reinforcement Learning with Audio-Text Semantic Rewards
arXiv paper on test-time reinforcement learning to improve automatic speech recognition robustness in noisy environments and diverse accents.
3/6/2026
Harnessing Synthetic Data from Generative AI for Statistical Inference
arXiv paper reviewing synthetic data generation from generative AI models for statistical inference. Covers validity and reliability of synthetic data usage.
3/6/2026
Ensembling Language Models with Sequential Monte Carlo
arXiv paper on ensemble methods for language models using Sequential Monte Carlo to aggregate predictions from multiple LLMs and prompting strategies.
3/6/2026
Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought
arXiv paper analyzing chain-of-thought reasoning in models, showing performative generation where models generate tokens without revealing internal beliefs.
3/6/2026
Path Planning for Masked Diffusion Model Sampling
Method for improving masked diffusion models with path planning to enable iterative refinement during token generation.
3/6/2026
BACE-RUL: A Bi-directional Adversarial Network with Covariate Encoding for Machine Remaining Useful Life Prediction
Bi-directional adversarial network with covariate encoding for predicting remaining useful life in machinery and equipment.
3/6/2026
ms-Mamba: Multi-scale Mamba for Time-Series Forecasting
Multi-scale Mamba architecture for time-series forecasting that processes inputs at multiple temporal scales using state space models.
3/6/2026
Learning Virtual Machine Scheduling in Cloud Computing through Language Agents
Language agents trained to solve VM scheduling (bin packing) in cloud computing with improved generalizability and interpretability over heuristic and learning-based baselines.
3/6/2026
VTool-R1: VLMs Learn to Think with Images via Reinforcement Learning on Multimodal Tool Use
VTool-R1 extends RL finetuning to vision-language models enabling multimodal reasoning with tool use and iterative image refinement during inference.
3/6/2026
Attribute-Efficient PAC Learning of Sparse Halfspaces with Constant Malicious Noise Rate
Theoretical work on attribute-efficient PAC learning of sparse halfspaces with robustness to constant-rate malicious noise in data.
3/6/2026
Continuous Chain of Thought Enables Parallel Exploration and Reasoning
CoT2 extends chain-of-thought reasoning in LLMs using continuous token representations instead of discrete sampling, with theoretical guarantees for logical reasoning tasks.
3/6/2026
FPGA-Enabled Machine Learning Applications in Earth Observation: A Systematic Review
Systematic review of FPGA-based ML deployment for real-time Earth observation processing on UAVs and satellite systems with bandwidth constraints.
3/6/2026
SPEED-RL: Faster Training of Reasoning Models via Online Curriculum Learning
SPEED-RL introduces an adaptive curriculum learning method for efficient RL training of reasoning LLMs by selectively sampling intermediate-difficulty prompts to maximize learning efficiency.
3/6/2026
Bures-Wasserstein Flow Matching for Graph Generation
Develops Bures-Wasserstein flow matching for graph generation using optimal transport-based probability paths for drug discovery and circuit design.
3/6/2026
From Bandit Regret to FDR Control: Online Selective Generation with Adversarial Feedback Unlocking
Proposes selective generation framework with FDR control for interactive generative systems using online learning with adversarial feedback.
3/6/2026
Structured Kolmogorov-Arnold Neural ODEs for Interpretable Learning and Symbolic Discovery of Nonlinear Dynamics
Presents SKANODEs integrating structured state-space modeling with Kolmogorov-Arnold networks for interpretable discovery of nonlinear dynamical systems.
3/6/2026
Parameter Stress Analysis in Reinforcement Learning: Applying Synaptic Filtering to Policy Networks
Analyzes policy network robustness using synaptic filtering to study parameter stress under internal and external adversarial perturbations.
3/6/2026
Overtone: Cyclic Patch Modulation for Clean, Efficient, and Flexible Physics Emulators
Proposes Overtone, a dynamic patch size modulation technique for transformer-based PDE surrogate models with improved efficiency and accuracy.
3/6/2026
Kernel Based Maximum Entropy Inverse Reinforcement Learning for Mean-Field Games
Develops kernel-based maximum entropy inverse reinforcement learning for mean-field games with nonlinear reward function inference from demonstrations.
3/6/2026
TIC-GRPO: Provable and Efficient Optimization for Reinforcement Learning from Human Feedback
Provides theoretical analysis and improvement of GRPO, a critic-free reinforcement learning algorithm for LLM fine-tuning via group-normalized rewards.
3/6/2026
In-Training Defenses against Emergent Misalignment in Language Models
Proposes in-training defenses against emergent misalignment in fine-tuned language models that induce harmful behaviors outside target domain.
3/6/2026
Multi-Agent Reinforcement Learning in Intelligent Transportation Systems: A Comprehensive Survey
Comprehensive survey of multi-agent reinforcement learning applications in intelligent transportation systems covering coordination and autonomous decision-making.
3/6/2026
TabStruct: Measuring Structural Fidelity of Tabular Data
Introduces TabStruct metric for evaluating structural fidelity of synthetic tabular data generators against real data causal structures.
3/6/2026
Complexity-Regularized Proximal Policy Optimization
Proposes complexity-regularized proximal policy optimization replacing entropy regularization with self-regulating complexity terms for policy gradient methods.
3/6/2026
Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer
Studies subliminal learning phenomenon where language models transfer hidden biases during distillation without explicit training signals.