3/12/2026
Adversarial Hubness Detector: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems
Detection method for hubness poisoning attacks in RAG systems where malicious items are frequently retrieved to inject harmful content.
3/12/2026
Defensive Refusal Bias: How Safety Alignment Fails Cyber Defenders
Study of safety alignment failures in LLMs for cybersecurity, where defensive refusal bias prevents assistance for legitimate defensive tasks.
3/12/2026
BrandFusion: A Multi-Agent Framework for Seamless Brand Integration in Text-to-Video Generation
Multi-agent framework for automatically embedding brand elements into text-to-video generation while preserving semantic fidelity to prompts.
3/12/2026
BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning
Bias-aware dynamic model merging approach using contrastive learning for multi-task learning under distribution shift.
3/12/2026
RACAS: Controlling Diverse Robots With a Single Agentic System
Agentic system enabling unified control of diverse robotic platforms through a single interface, bridging low-level APIs to high-level autonomous behavior.
3/12/2026
ResearchEnvBench: Benchmarking Agents on Environment Synthesis for Research Code Execution
Benchmark for evaluating autonomous agents on environment synthesis for research code execution, addressing complex dependencies and configuration challenges.
3/12/2026
SeDa: A Unified System for Dataset Discovery and Multi-Entity Augmented Semantic Exploration
Unified framework integrating 7.6M datasets from 200+ platforms for cross-source data discovery, semantic annotation, and navigation.
3/12/2026
A Systematic Comparison of Training Objectives for Out-of-Distribution Detection in Image Classification
Systematic comparison of training objectives (Cross-Entropy, Prototype, Triplet, AP Loss) for out-of-distribution detection in image classification.
3/12/2026
Alignment-Process-Outcome: Rethinking How AIs and Humans Collaborate
Framework for understanding collaboration between AI systems and humans, examining alignment, process structure, and outcome quality relationships.
3/12/2026
SiliconMind-V1: Multi-Agent Distillation and Debug-Reasoning Workflows for Verilog Code Generation
Multi-agent framework for Verilog code generation using LLM distillation and debug-reasoning workflows, addressing cost and functional correctness without relying on commercial models.
3/12/2026
Alignment as Iatrogenesis: Pastoral Power, Collective Pathology, and the Structural Limits of Monolingual Safety Evaluation
Theoretical analysis of LLM safety alignment as creating behavioral pathologies using Foucauldian framework.
3/12/2026
Beyond Relevance: On the Relationship Between Retrieval and RAG Information Coverage
Study examining relationship between retrieval quality and generation effectiveness in RAG systems.
3/12/2026
Fish Audio S2 Technical Report
Open-source text-to-speech system with multi-speaker generation and natural-language instruction control.
3/12/2026
PathoScribe: Transforming Pathology Data into a Living Library with a Unified LLM-Driven Framework for Semantic Retrieval and Clinical Integration
PathoScribe LLM-driven framework transforms pathology narrative reports into searchable semantic library for clinical integration and knowledge retrieval.
3/12/2026
PlayWorld: Learning Robot World Models from Autonomous Play
PlayWorld autonomous pipeline for learning robot world models through self-play using action-conditioned video models for manipulation tasks.
3/12/2026
VIVID-Med: LLM-Supervised Structured Pretraining for Deployable Medical ViTs
VIVID-Med framework uses frozen LLM as semantic teacher for structured vision-language pretraining of medical image analysis models.
3/12/2026
SPAARS: Safer RL Policy Alignment through Abstract Exploration and Refined Exploitation of Action Space
SPAARS method for safer offline-to-online RL in robotics using action space abstraction to constrain exploration within behavioral support.
3/12/2026
MM-tau-p$^2$: Persona-Adaptive Prompting for Robust Multi-Modal Agent Evaluation in Dual-Control Settings
Evaluation framework for LLM-powered multimodal agents in customer service with persona-adaptive prompting for dual-control settings.
3/12/2026
AutoViVQA: A Large-Scale Automatically Constructed Dataset for Vietnamese Visual Question Answering
Large-scale Vietnamese Visual Question Answering dataset automatically constructed using pre-trained transformers and language models.
3/12/2026
Ego: Embedding-Guided Personalization of Vision-Language Models
Embedding-guided personalization method for vision-language models enabling user-specific customization without additional training stages.
3/12/2026
MA-EgoQA: Question Answering over Egocentric Videos from Multiple Embodied Agents
Dataset and QA framework for multi-agent egocentric video understanding enabling communication between humans and embodied AI agents.
3/12/2026
Explainable LLM Unlearning Through Reasoning
Research on explainable LLM unlearning using reasoning-based approach to remove undesirable knowledge while preserving general capabilities.
3/12/2026
MoE-SpAc: Efficient MoE Inference Based on Speculative Activation Utility in Heterogeneous Edge Scenarios
MoE-SpAc uses speculative decoding for efficient Mixture-of-Experts inference on edge devices addressing memory constraints.
3/12/2026
Personalized Group Relative Policy Optimization for Heterogenous Preference Alignment
Group Relative Policy Optimization extended for personalized alignment with heterogeneous user preferences in LLM post-training.
3/12/2026
LWM-Temporal: Sparse Spatio-Temporal Attention for Wireless Channel Representation Learning
Foundation model for wireless channel representation learning using sparse spatio-temporal attention in angle-delay-time domain.
3/12/2026
Gated Adaptation for Continual Learning in Human Activity Recognition
Continual learning method using gated adaptation for human activity recognition in IoT wearable sensors addressing catastrophic forgetting.
3/12/2026
Revisiting Sharpness-Aware Minimization: A More Faithful and Effective Implementation
Faithful implementation of Sharpness-Aware Minimization optimizer correcting approximations in standard SAM to improve LLM generalization.
3/12/2026
InFusionLayer: a CFA-based ensemble tool to generate new classifiers for learning and modeling
Python tool implementing Combinatorial Fusion Analysis for ensemble learning and classifier generation combining multiple models.
3/12/2026
Cluster-Aware Attention-Based Deep Reinforcement Learning for Pickup and Delivery Problems
Deep reinforcement learning approach using cluster-aware attention for vehicle routing pickup and delivery problems with spatial constraints.
3/12/2026
Training Language Models via Neural Cellular Automata
Research proposes neural cellular automata for generating synthetic non-linguistic training data as alternative to natural language for LLM pretraining.
3/12/2026
HTMuon: Improving Muon via Heavy-Tailed Spectral Correction
HTMuon optimizer improves LLM training by correcting heavy-tailed weight spectra issues in Muon algorithm using spectral theory.
3/12/2026
Improving Search Agent with One Line of Code
Research on Tool-based Agentic RL identifies Importance Sampling Distribution Drift causing training instability in search agents using external tools.
3/12/2026
Dissecting Chronos: Sparse Autoencoders Reveal Causal Feature Hierarchies in Time Series Foundation Models
Sparse autoencoders applied to Chronos time series foundation model reveal causal feature hierarchies through ablation studies.
3/12/2026
Digging Deeper: Learning Multi-Level Concept Hierarchies
Hierarchical concept embedding models learn multi-level concept hierarchies with coarse annotations for interpretable predictions.
3/12/2026
KernelSkill: A Multi-Agent Framework for GPU Kernel Optimization
KernelSkill multi-agent framework uses LLMs as interpretable optimization agents for GPU kernel development and tuning.
3/12/2026
ES-dLLM: Efficient Inference for Diffusion Large Language Models by Early-Skipping
ES-dLLM optimizes diffusion language model inference through early-skipping of redundant computations in intermediate layers.
3/12/2026
A Survey of Weight Space Learning: Understanding, Representation, and Generation
Survey of weight space learning covering structure, symmetries, and generation of neural network weights as structured objects.
3/12/2026
Equivariant Asynchronous Diffusion: An Adaptive Denoising Schedule for Accelerated Molecular Conformation Generation
Equivariant asynchronous diffusion model with adaptive denoising schedule for 3D molecular conformation generation.
3/12/2026
Rethinking Adam for Time Series Forecasting: A Simple Heuristic to Improve Optimization under Distribution Shifts
Modified Adam optimizer for time series forecasting addresses non-stationarity and distribution shifts in adaptive learning.
3/12/2026
CLIPO: Contrastive Learning in Policy Optimization Generalizes RLVR
CLIPO contrastive learning framework for LLM reasoning improves upon RLVR by training on intermediate reasoning correctness.
3/12/2026
Lost in the Middle at Birth: An Exact Theory of Transformer Position Bias
Theoretical analysis showing lost-in-middle phenomenon in transformers emerges at initialization as geometric property before training.
3/12/2026
A neural operator for predicting vibration frequency response curves from limited data
Neural operator approach for predicting vibration frequency response from limited data in mechanical engineering design.
3/12/2026
Mashup Learning: Faster Finetuning by Remixing Past Checkpoints
Mashup Learning reuses and remixes past model checkpoints to accelerate LLM finetuning on domain-specific tasks.
3/12/2026
ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning
ReMix applies reinforcement learning to route inputs across specialized LoRA adapters during LLM finetuning for parameter efficiency.
3/12/2026
Actor-Accelerated Policy Dual Averaging for Reinforcement Learning in Continuous Action Spaces
Actor-accelerated policy dual averaging enables reinforcement learning in continuous action spaces with approximate value functions.
3/12/2026
Rethinking the Harmonic Loss via Non-Euclidean Distance Layers
Harmonic loss function using non-Euclidean distance layers addresses interpretability and training inefficiencies in deep neural networks.
3/12/2026
SiMPO: Measure Matching for Online Diffusion Reinforcement Learning
SiMPO framework generalizes policy optimization for diffusion-based reinforcement learning with monotonic reweighting functions.
3/12/2026
Improving TabPFN's Synthetic Data Generation by Integrating Causal Structure
TabPFN foundation model for tabular data improved by integrating causal structure into synthetic data generation process.
3/12/2026
Discovery of a Hematopoietic Manifold in scGPT Yields a Method for Extracting Performant Algorithms from Biological Foundation Model Internals
Mechanistic interpretability applied to scGPT foundation model extracts compact hematopoietic algorithm from model internals with biological validation.
3/12/2026
Robust Post-Training for Generative Recommenders: Why Exponential Reward-Weighted SFT Outperforms RLHF
Post-training method for generative recommender systems using exponential reward-weighted supervised fine-tuning. Compares with RLHF and offline RL approaches for production systems.