3/9/2026
Prediction-Powered Conditional Inference
Develops prediction-powered conditional inference method combining localization and prediction-based variance reduction for scarce labeled data.
3/9/2026
RACAS: Controlling Diverse Robots With a Single Agentic System
Presents RACAS, an agentic system for controlling diverse robotic platforms with unified API and autonomous behavior pipeline.
3/9/2026
Making Reconstruction FID Predictive of Diffusion Generation FID
Introduces interpolated FID metric to improve correlation between VAE reconstruction and diffusion model generation quality.
3/9/2026
When Rubrics Fail: Error Enumeration as Reward in Reference-Free RL Post-Training for Virtual Try-On
Proposes error enumeration as reward signal for RL post-training in reference-free settings without ideal answer references.
3/9/2026
Parallelization Strategies for Dense LLM Deployment: Navigating Through Application-Specific Tradeoffs and Bottlenecks
Comprehensive analysis of parallelization strategies for dense LLM deployment, evaluating tradeoffs between tensor/pipeline/sequence parallelism.
3/9/2026
Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models
Mechanistic analysis of LLM safety mechanisms revealing decoupling between harmfulness recognition and refusal via disentangled geometry.
3/9/2026
ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning
LLMs trained via RL to self-reflect and correct generated code without external oracles, improving complex algorithmic task performance.
3/9/2026
ROSE: Reordered SparseGPT for More Accurate One-Shot Large Language Models Pruning
Improved one-shot LLM pruning using optimal weight reordering instead of predefined order, advancing SparseGPT methodology.
3/9/2026
Addressing the Ecological Fallacy in Larger LMs with Human Context
Addresses ecological fallacy in language models by incorporating author context through specialized LM pretraining tasks.
3/9/2026
Implicit Style Conditioning: A Structured Style-Rewrite Framework for Low-Resource Character Modeling
Framework for fine-tuning small language models with stylized personas using structured style-rewriting to improve character consistency.
3/9/2026
Who We Are, Where We Are: Mental Health at the Intersection of Person, Situation, and Large Language Models
Interpretable models using LLMs to predict mental health and well-being from longitudinal social media data by integrating psychological traits.
3/9/2026
TADPO: Reinforcement Learning Goes Off-road
Reinforcement learning approach for off-road autonomous driving handling unmapped terrain, variable dynamics, and long-horizon planning.
3/9/2026
Diffusion Language Models Are Natively Length-Aware
Diffusion Language Models adapt generation length dynamically, reducing computational waste on short responses in reasoning tasks.
3/9/2026
Efficient Vector Search in the Wild: One Model for Multi-K Queries
Method for efficient vector search that generalizes across multiple K values in top-K retrieval without retraining, improving serving performance.
3/9/2026
Learning Where the Physics Is: Probabilistic Adaptive Sampling for Stiff PDEs
Adaptive sampling method using Gaussian Mixture Models to improve Physics-Informed Neural Networks training on stiff PDEs.
3/9/2026
SAHOO: Safeguarded Alignment for High-Order Optimization Objectives in Recursive Self-Improvement
Framework for monitoring and preventing alignment drift in recursive self-improvement systems using multi-signal detection and constraint preservation.
3/9/2026
MoEless: Efficient MoE LLM Serving via Serverless Computing
Serverless deployment system for efficient serving of Mixture-of-Experts LLMs by optimizing sparse activation patterns.
3/9/2026
Dynamic Chunking Diffusion Transformer
Diffusion Transformer variant with dynamic token chunking that adapts compute allocation based on image content detail and denoising stages.
3/9/2026
Kinetic-based regularization: Learning spatial derivatives and PDE applications
Kinetic-based regularization extension for learning spatial derivatives from noisy data with provable accuracy for PDE applications.
3/9/2026
Talk Freely, Execute Strictly: Schema-Gated Agentic AI for Flexible and Reproducible Scientific Workflows
Framework enabling LLMs to execute scientific workflows with schema-gated constraints ensuring determinism, provenance, and governance.
3/9/2026
Efficient, Property-Aligned Fan-Out Retrieval via RL-Compiled Diffusion
RL-based method for retrieving diverse, property-aligned result sets using diffusion models for set-valued retrieval objectives.
3/9/2026
A Reference Architecture of Reinforcement Learning Frameworks
Reference architecture framework analyzing 18 RL implementations to establish common patterns and standardization for RL frameworks.
3/9/2026
CLoPA: Continual Low Parameter Adaptation of Interactive Segmentation for Medical Image Annotation
Continual learning strategy for online adaptation of interactive segmentation models in medical imaging with low-parameter updates.
3/9/2026
Certified and accurate computation of function space norms of deep neural networks
Method for computing certified bounds on function space norms of deep neural networks applied to PDE solutions.
3/9/2026
Semantics-Aware Caching for Concept Learning
Optimization technique using semantic-aware caching to improve instance retrieval efficiency in concept learning on knowledge bases.
3/9/2026
Boosting deep Reinforcement Learning using pretraining with Logical Options
Two-stage hybrid framework combining logical options with deep reinforcement learning to improve agent alignment and prevent over-exploitation of early reward signals.
3/9/2026
SCOPE: Scene-Contextualized Incremental Few-Shot 3D Segmentation
SCOPE incremental few-shot 3D point cloud segmentation addressing catastrophic forgetting by leveraging unlabeled background scenes in sparse supervision settings.
3/9/2026
BEVLM: Distilling Semantic Knowledge from LLMs into Bird's-Eye View Representations
BEVLM distills semantic knowledge from LLMs into bird's-eye view representations for autonomous driving with improved spatial consistency and reduced computation.
3/9/2026
Predictive Coding Networks and Inference Learning: Tutorial and Survey
Tutorial and survey on predictive coding networks based on neuroscientific framework viewing brain as hierarchical Bayesian inference model minimizing prediction errors.
3/9/2026
PACE: Marrying generalization in PArameter-efficient fine-tuning with Consistency rEgularization
PACE combines parameter-efficient fine-tuning with consistency regularization to improve generalization by reducing gradient norms during transformer adaptation to downstream tasks.
3/9/2026
FragFM: Hierarchical Framework for Efficient Molecule Generation via Fragment-Level Discrete Flow Matching
FragFM hierarchical framework using fragment-level discrete flow matching and coarse-to-fine autoencoders for efficient scalable molecular graph generation.
3/9/2026
Aligning Compound AI Systems via System-level DPO
System-level DPO method for aligning compound AI systems with multiple interacting components including LLMs, foundation models, and external tools to human preferences.
3/9/2026
CAPS: Context-Aware Priority Sampling for Enhanced Imitation Learning in Autonomous Driving
Context-Aware Priority Sampling using VQ-VAEs to improve data efficiency and handle imbalanced datasets in imitation learning for autonomous driving systems.
3/9/2026
From Tokenizer Bias to Backbone Capability: A Controlled Study of LLMs for Time Series Forecasting
Controlled study examining how LLM tokenizer bias and backbone capability affect time series forecasting performance using pre-trained language models as backbone.
3/9/2026
Federated Learning: A Survey on Privacy-Preserving Collaborative Intelligence
Federated Learning survey covering privacy-preserving distributed machine learning enabling multiple clients to collaboratively train models without centralizing sensitive data.
3/9/2026
FourierSpecNet: Neural Collision Operator Approximation Inspired by the Fourier Spectral Method for Solving the Boltzmann Equation
FourierSpecNet hybrid framework combining Fourier spectral methods with neural networks to approximate collision operators for solving the Boltzmann equation efficiently.
3/9/2026
Merging Memory and Space: A State Space Neural Operator
State Space Neural Operator for learning solution operators of time-dependent PDEs using structured state space models with adaptive damping and learnable frequency modulation.
3/9/2026
TIC-GRPO: Provable and Efficient Optimization for Reinforcement Learning from Human Feedback
arXiv paper providing theoretical analysis of GRPO (Group Relative Policy Optimization) for LLM fine-tuning from human feedback.
3/9/2026
Characterizing Evolution in Expectation-Maximization Estimates for Overspecified Mixed Linear Regression
arXiv paper analyzing EM algorithm behavior under model misspecification in mixture models with excess components.
3/9/2026
A Geometric Perspective on the Difficulties of Learning GNN-based SAT Solvers
arXiv paper analyzing GNN-based SAT solvers through graph Ricci curvature geometric perspective to explain performance degradation.
3/9/2026
One Model for All Tasks: Leveraging Efficient World Models in Multi-Task Planning
arXiv paper on efficient world models for heterogeneous multi-task planning, addressing gradient conflicts and plasticity loss.
3/9/2026
Performance Assessment Strategies for Language Model Applications in Healthcare
arXiv paper on assessing performance of language model applications in healthcare, addressing evaluation methodology.
3/9/2026
Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check
arXiv paper introducing Answer-Then-Check safety alignment method to defend LLMs against jailbreak attacks using reasoning.
3/9/2026
C^2Prompt: Class-aware Client Knowledge Interaction for Federated Continual Learning
arXiv paper on prompt-based federated continual learning addressing class-wise and temporal forgetting across distributed clients.
3/9/2026
Planner Aware Path Learning in Diffusion Language Models Training
arXiv paper on training diffusion language models with planner-aware path learning to optimize generation strategies.
3/9/2026
Diffusion Alignment as Variational Expectation-Maximization
arXiv paper formulating diffusion model alignment as variational EM to reduce reward over-optimization and mode collapse.
3/9/2026
Decoding Partial Differential Equations: Cross-Modal Adaptation of Decoder-only Models to PDEs
arXiv paper on adapting decoder-only LLMs to partial differential equations via cross-modal learning for scientific machine learning.
3/9/2026
KLASS: KL-Guided Fast Inference in Masked Diffusion Models
arXiv paper introducing KLASS sampling method for masked diffusion models using token-level KL divergence to accelerate inference.
3/9/2026
Diffusion Fine-Tuning via Reparameterized Policy Gradient of the Soft Q-Function
SQDF applies soft Q-function RL to fine-tune diffusion models with KL regularization, mitigating reward over-optimization.
3/9/2026
Whatever Remains Must Be True: Filtering Drives Reasoning in LLMs, Shaping Diversity
Analyzes diversity loss in RL-trained LLMs caused by mode-seeking reverse KL; proposes forward KL filtering for reasoning tasks.