2/13/2026
General and Efficient Steering of Unconditional Diffusion
Efficient steering method for unconditional diffusion models enabling controllable generation without gradient guidance or retraining.
2/13/2026
Can We Really Learn One Representation to Optimize All Rewards?
Investigates whether a single learned representation can optimize multiple reward functions in RL, studying representation learning for multi-objective policies.
2/13/2026
CADET: Context-Conditioned Ads CTR Prediction With a Decoder-Only Transformer
CADET applies decoder-only transformer architecture to CTR prediction in ads systems, addressing challenges of contextual post-scoring constraints.
2/13/2026
Hierarchical Concept Embedding & Pursuit for Interpretable Image Classification
Interpretable image classification using hierarchical concept embeddings derived from vision-language models for sparse concept recovery.
2/13/2026
External Division of Two Bregman Proximity Operators for Poisson Inverse Problems
Novel sparse recovery method for Poisson inverse problems using Bregman proximity operators and NoLips algorithm.
2/13/2026
Exploring Multiple High-Scoring Subspaces in Generative Flow Networks
Research on Generative Flow Networks: methods for biasing GFlowNets toward high-reward solutions in combinatorial optimization.
2/13/2026
Partial GFlowNet: Accelerating Convergence in Large State Spaces via Strategic Partitioning
Partial GFlowNet accelerates convergence in large state spaces via planner-based partitioning, improving generative flow network scalability.
2/13/2026
RooflineBench: A Benchmarking Framework for On-Device LLMs via Roofline Analysis
RooflineBench framework benchmarks small language models on edge hardware using roofline analysis for performance characterization across architectures.
2/13/2026
Unifying Stable Optimization and Reference Regularization in RLHF
Unified framework addressing reward hacking and optimization instability in RLHF by combining KL regularization and policy ratio clipping.
2/13/2026
Adaptive Milestone Reward for GUI Agents
Adaptive Milestone Reward addresses temporal credit assignment in RL-trained GUI agents by balancing outcome and process reward with adaptive thresholds.
2/13/2026
PASCAL: A Phase-Aware Scheduling Algorithm for Serving Reasoning-based Large Language Models
PASCAL phase-aware scheduling algorithm optimizes serving of reasoning-based LLMs by distinguishing reasoning and answering phases for improved latency.
2/13/2026
AltTS: A Dual-Path Framework with Alternating Optimization for Multivariate Time Series Forecasting
AltTS dual-path framework with alternating optimization separates autoregressive dynamics from cross-dimension interactions in multivariate time series forecasting.
2/13/2026
Krause Synchronization Transformers
Krause Attention mechanism prevents representation collapse in transformers by decoupling softmax normalization inspired by bounded-confidence dynamics.
2/13/2026
Real-Time Proactive Anomaly Detection via Forward and Backward Forecast Modeling
Proactive anomaly detection using forward and backward forecast modeling for early warning signals in industrial, financial, cybersecurity applications.
2/13/2026
Native Reasoning Models: Training Language Models to Reason on Unverifiable Data
Native Reasoning Models trains language models to reason without external verifiers or fully-annotated data, extending RLVR paradigm to unverifiable domains.
2/13/2026
TS-Memory: Plug-and-Play Memory for Time Series Foundation Models
TS-Memory adds plug-and-play memory module to time series foundation models for efficient adaptation under distribution shift without catastrophic forgetting.
2/13/2026
The Implicit Bias of Steepest Descent with Mini-batch Stochastic Gradient
Studies implicit bias of mini-batch stochastic steepest descent in multiclass classification under various norm geometries.
2/13/2026
Brain4FMs: A Benchmark of Foundation Models for Electrical Brain Signal
Benchmark evaluating foundation models pretrained on brain electrical signals for EEG and intracranial recordings in neuroscience.
2/13/2026
Gradient Compression May Hurt Generalization: A Remedy by Synthetic Data Guided Sharpness Aware Minimization
Analyzes gradient compression effects on loss landscapes in federated learning, proposing sharpness-aware minimization remedy for generalization.
2/13/2026
Learn from Your Mistakes: Self-Correcting Masked Diffusion Models
Framework enabling masked diffusion models to perform token correction after unmasking, reducing error accumulation in parallel generation.
2/13/2026
SkillRater: Untangling Capabilities in Multimodal Data
SkillRater decomposes data quality into multidimensional capabilities rather than single scores, improving data curation for model training.
2/13/2026
How Well Do Large-Scale Chemical Language Models Transfer to Downstream Tasks?
Evaluates transfer learning performance of large-scale chemical language models across downstream molecular property prediction tasks.
2/13/2026
TreeGrad-Ranker: Feature Ranking via $O(L)$-Time Gradients for Decision Trees
TreeGrad-Ranker uses probabilistic values to rank features in decision trees for local prediction explanation via efficient gradient computation.
2/13/2026
ArGEnT: Arbitrary Geometry-encoded Transformer for Operator Learning
ArGEnT transformer learns solution operators for physical systems with complex geometries, enabling surrogate modeling across varying parametric settings.
2/13/2026
GP2F: Cross-Domain Graph Prompting with Adaptive Fusion of Pre-trained Graph Neural Networks
Graph prompt learning framework adapting pre-trained GNNs across domains via adaptive fusion and cross-domain knowledge transfer.
2/13/2026
TIP: Resisting Gradient Inversion via Targeted Interpretable Perturbation in Federated Learning
Defense against gradient inversion attacks in federated learning using targeted interpretable perturbation preserving model utility.
2/13/2026
Both Topology and Text Matter: Revisiting LLM-guided Out-of-Distribution Detection on Text-attributed Graphs
LLM-guided out-of-distribution detection for text-attributed graphs combining topology and text features for unseen patterns.
2/13/2026
DRACO: a Cross-Domain Benchmark for Deep Research Accuracy, Completeness, and Objectivity
Benchmark of complex deep research tasks across 10 domains from real-world usage patterns, measuring accuracy, completeness, objectivity.
2/13/2026
ANML: Attribution-Native Machine Learning with Guaranteed Robustness
Framework weighting training samples by quality factors including gradient consistency and verification status for specialized expert data.
2/13/2026
SpiralFormer: Looped Transformers Can Learn Hierarchical Dependencies via Multi-Resolution Recursion
Recursive Transformer architecture learning hierarchical dependencies through multi-resolution recursion with shared layers.
2/13/2026
TabSieve: Explicit In-Table Evidence Selection for Tabular Prediction
Framework for tabular prediction using explicit in-table evidence selection, making row context auditable and interpretable.
2/13/2026
DICE: Diffusion Large Language Models Excel at Generating CUDA Kernels
Diffusion LLM approach for CUDA kernel code generation leveraging parallel token generation and non-sequential refinement.
2/13/2026
Dopamine: Brain Modes, Not Brains
Parameter-efficient fine-tuning method viewing adaptation as neuromodulation-inspired mode selection and rescaling of pretrained computations.
2/13/2026
U-Former ODE: Fast Probabilistic Forecasting of Irregular Time Series
U-Former ODE architecture for fast probabilistic forecasting of irregularly sampled time series using neural differential equations.
2/13/2026
TUBO: A Tailored ML Framework for Reliable Network Traffic Forecasting
ML framework for reliable network traffic forecasting using deep learning adapted for traffic characteristics.
2/13/2026
MUSE: Multi-Tenant Model Serving With Seamless Model Updates
Multi-tenant model serving system handling seamless model updates with dynamic decision threshold management.
2/13/2026
Temperature as a Meta-Policy: Adaptive Temperature in LLM Reinforcement Learning
Framework treating temperature as adaptive meta-policy in LLM reinforcement learning to improve exploration-exploitation tradeoff.
2/13/2026
Safe Fairness Guarantees Without Demographics in Classification: Spectral Uncertainty Set Perspective
Method for fair classification without demographic information using spectral uncertainty sets.
2/13/2026
Evaluating LLM Safety Under Repeated Inference via Accelerated Prompt Stress Testing
Benchmark for evaluating LLM safety under repeated inference via prompt stress testing, addressing consistency failures in deployment.
2/13/2026
Temporal Difference Learning with Constrained Initial Representations
Enhances off-policy RL sample efficiency by constraining initial representations to address distribution shift and stabilize training.
2/13/2026
TopoFair: Linking Topological Bias to Fairness in Link Prediction Benchmarks
TopoFair analyzes topological biases in graph link prediction benchmarks beyond homophily to improve fairness in recommendations.
2/13/2026
Deep Kernel Fusion for Transformers
DeepFusionKernel optimizes agentic LLM inference by fusing SwiGLU blocks, reducing memory bandwidth bottlenecks by 9-13% on A100/H100.
2/13/2026
In-Context Function Learning in Large Language Models
Analyzes in-context learning in LLMs through Gaussian Process lens, studying function learning from few demonstrations at inference.
2/13/2026
Where Bits Matter in World Model Planning: A Paired Mixed-Bit Study for Efficient Spatial Reasoning
Studies mixed-bit quantization strategies for world model planning, analyzing where precision allocation matters most.
2/13/2026
Mitigating Mismatch within Reference-based Preference Optimization
Addresses reference policy mismatch in DPO by handling pessimistic preference pairs where reference model prefers rejected responses.
2/13/2026
Extending Puzzle for Mixture-of-Experts Reasoning Models with Application to GPT-OSS Acceleration
Extends Puzzle NAS framework to optimize gpt-oss-120B into 88B model using MoE pruning and attention optimization for inference acceleration.
2/13/2026
Temporally Unified Adversarial Perturbations for Time Series Forecasting
Proposes temporally consistent adversarial perturbations for time series forecasting models to improve robustness against attacks.
2/13/2026
Using predictive multiplicity to measure individual performance within the AI Act
Addresses predictive multiplicity in AI decision support systems, measuring individual performance disagreement under the AI Act framework.
2/13/2026
Towards Performance-Enhanced Model-Contrastive Federated Learning using Historical Information in Heterogeneous Scenarios
PMFL framework improves federated learning in heterogeneous scenarios using model-contrastive learning with historical training information.
2/13/2026
Are Two LLMs Better Than One? A Student-Teacher Dual-Head LLMs Architecture for Pharmaceutical Content Optimization
LRBTC system uses dual LLM and VLM architecture for automated quality control of pharmaceutical content covering regulatory, brand, and technical compliance.