4/14/2026
SCITUNE: Aligning Large Language Models with Human-Curated Scientific Multimodal Instructions
SciTune framework aligns LLMs with scientific domain knowledge through instruction fine-tuning on multimodal scientific publication data.
4/14/2026
MM-LIMA: Less Is More for Alignment in Multi-Modal Datasets
MM-LIMA demonstrates multimodal LLM fine-tuning achieves strong results with only 200 high-quality instruction examples, reducing data requirements.
4/14/2026
CROP: Conservative Reward for Model-based Offline Policy Optimization
Proposes CROP, a model-based offline reinforcement learning method addressing distribution shift through conservative reward estimation.
4/14/2026
An Iterative Utility Judgment Framework Inspired by Philosophical Relevance via LLMs
Framework using LLMs with philosophical relevance concepts to improve utility-based result ranking in retrieval-augmented generation systems.
4/14/2026
MegaFake: A Theory-Driven Dataset of Fake News Generated by Large Language Models
MegaFake dataset of LLM-generated fake news for studying mechanisms of misinformation generation and detection methods.
4/14/2026
Deep Optimizer States: Towards Scalable Training of Transformer Models Using Interleaved Offloading
Deep Optimizer States method enables scalable training of transformer models using interleaved offloading to overcome memory constraints.
4/14/2026
PoTable: Towards Systematic Thinking via Plan-then-Execute Stage Reasoning on Tables
PoTable framework improves table reasoning in LLMs using plan-then-execute reasoning stages for systematic thinking.
4/14/2026
WebLLM: A High-Performance In-Browser LLM Inference Engine
WebLLM inference engine enabling high-performance LLM execution directly in web browsers for on-device deployment without server GPUs.
4/14/2026
HumanVBench: Probing Human-Centric Video Understanding in MLLMs with Automatically Synthesized Benchmarks
HumanVBench benchmark for evaluating human-centric video understanding in multimodal large language models with 16 fine-grained tasks.
4/14/2026
Influencing Humans to Conform to Preference Models for RLHF
Three human studies examining whether humans can be influenced to conform to preference models used in RLHF algorithms for LLMs.
4/14/2026
Curriculum-based Sample Efficient Reinforcement Learning for Robust Stabilization of a Quadrotor
Novel curriculum learning approach for sample-efficient reinforcement learning applied to quadrotor stabilization control.
4/14/2026
Integrating Semi-Supervised and Active Learning for Semantic Segmentation
Combines semi-supervised and active learning for semantic segmentation to reduce manual annotation costs and improve model performance.
4/14/2026
Large Language Models Can Help Mitigate Barren Plateaus in Quantum Neural Networks
Proposes using LLMs to help mitigate barren plateaus in quantum neural network training through adaptive parameter initialization.
4/14/2026
If an LLM Were a Character, Would It Know Its Own Story? Evaluating Lifelong Learning in LLMs
Study evaluating emergent lifelong learning behaviors in LLMs during multi-turn interactions, proposing new evaluation benchmarks for character-like consistency.
4/14/2026
TARAC: Mitigating Hallucination in LVLMs via Temporal Attention Real-time Accumulative Connection
TARAC method addresses hallucinations in vision-language models by improving temporal attention mechanisms during generation without extensive retraining.
4/14/2026
Optimizing Large Language Models: Metrics, Energy Efficiency, and Case Study Insights
Research on energy-efficient optimization techniques for LLM deployment, including quantization and local inference strategies to reduce carbon emissions.
4/14/2026
Not All Rollouts are Useful: Down-Sampling Rollouts in LLM Reinforcement Learning
PODS decouples rollout generation from policy updates in LLM RL, addressing compute asymmetry through down-sampling.
4/14/2026
LOOPE: Learnable Optimal Patch Order in Positional Embeddings for Vision Transformers
LOOPE method learns optimal patch ordering in Vision Transformer positional embeddings for improved spatial information encoding.
4/14/2026
Putting the Value Back in RL: Better Test-Time Scaling by Unifying LLM Reasoners With Verifiers
RL^V framework unifies LLM reasoners with verifiers using value functions for improved test-time compute scaling during reasoning.
4/14/2026
Variational Visual Question Answering for Uncertainty-Aware Selective Prediction
Bayesian approach for Vision Language Models to reduce hallucinations and overconfidence in VQA through selective prediction.
4/14/2026
TokUR: Token-Level Uncertainty Estimation for Large Language Model Reasoning
TokUR enables LLMs to self-assess uncertainty at token-level for improved reasoning and response reliability in multi-step tasks.
4/14/2026
SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence
SpatialScore: comprehensive benchmark for evaluating spatial intelligence of multimodal LLMs with data-driven and agent-based assessment approaches.
4/14/2026
GoT-R1: Unleashing Reasoning Capability of MLLM for Visual Generation with Reinforcement Learning
GoT-R1: reinforcement learning framework enhancing multimodal LLM reasoning for complex visual generation with precise spatial relationships and attributes.
4/14/2026
Tuning Language Models for Robust Prediction of Diverse User Behaviors
Fine-tuning approach for LLMs to predict diverse user behaviors, addressing overfitting to frequent behaviors while capturing long-tailed behavior distribution.
4/14/2026
Learning World Models for Interactive Video Generation
World models for interactive video generation with action conditioning and autoregressive decoding to support planning and future prediction.
4/14/2026
LLM-based Realistic Safety-Critical Driving Video Generation
Framework using LLMs for few-shot code generation to create safety-critical driving scenarios in CARLA simulator for autonomous driving evaluation.
4/14/2026
Large Language Model as An Operator: An Experience-Driven Solution for Distribution Network Voltage Control
LLM-based autonomous agent for power system voltage control, using experience-driven learning to generate dispatch strategies in distribution networks.
4/14/2026
Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training
Data Mixing Agent: LLM-based method to automatically re-weight training data domains during continual pre-training, preventing catastrophic forgetting.
4/14/2026
PRIX: Learning to Plan from Raw Pixels for End-to-End Autonomous Driving
PRIX: efficient end-to-end autonomous driving model planning from raw camera pixels without LiDAR, reducing model size and computational requirements.
4/14/2026
Modular Delta Merging with Orthogonal Constraints: A Scalable Framework for Continual and Reversible Model Composition
MDM-OC: framework for scalable, reversible model composition enabling continual learning without task interference or catastrophic forgetting.
4/14/2026
Teaching the Teacher: The Role of Teacher-Student Smoothness Alignment in Genetic Programming-based Symbolic Distillation
Genetic programming approach for symbolic distillation of neural networks, using teacher-student smoothness alignment to improve explainable AI model accuracy.
4/14/2026
Reliable Evaluation Protocol for Low-Precision Retrieval
Protocol for reliable evaluation of low-precision retrieval systems, addressing spurious ties and variability in relevance scoring with reduced numerical precision.
4/14/2026
Echoes of Automation: The Increasing Use of LLMs in Newsmaking
Analysis of LLM use in newsmaking across 40,000+ articles using AI-text detectors, showing increased GenAI adoption in local and college media.
4/14/2026
Proximal Supervised Fine-Tuning
Proximal SFT: supervised fine-tuning method using trust-region constraints to prevent capability deterioration when adapting foundation models to new tasks.
4/14/2026
Multi-Model Synthetic Training for Mission-Critical Small Language Models
LLM-based synthetic training reduces maritime domain model costs 261x by using LLMs as teachers for small language model training.
4/14/2026
FS-DFM: Fast and Accurate Long Text Generation with Few-Step Diffusion Language Models
FS-DFM enables fast long text generation using few-step diffusion language models with parallel position generation.
4/14/2026
StyleBench: Evaluating thinking styles in Large Language Models
StyleBench evaluates trade-offs between structured reasoning styles and efficiency/robustness in LLM inference.
4/14/2026
Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards
Position paper analyzing measurement gaps in reinforcement learning with verifiable rewards for LLMs on structured tasks.
4/14/2026
SecureVibeBench: Evaluating Secure Coding Capabilities of Code Agents with Realistic Vulnerability Scenarios
SecureVibeBench evaluates code generation security of LLM-powered code agents against realistic vulnerability scenarios.
4/14/2026
A Mathematical Explanation of Transformers
Mathematical framework interpreting Transformers as discretizations of integro-differential equations.
4/14/2026
EDUMATH: Generating Standards-aligned Educational Math Word Problems
LLM-based system for generating standards-aligned math word problems customized to student interests and ability levels.
4/14/2026
HiPRAG: Hierarchical Process Rewards for Efficient Agentic Retrieval Augmented Generation
HiPRAG uses hierarchical process rewards to improve agentic RAG efficiency, reducing over-search and under-search behaviors.
4/14/2026
Design Principles for Sequence Models via Coefficient Dynamics
Unified framework analyzing sequence models (Transformers, SSMs, gated RNNs) through coefficient dynamics lens.
4/14/2026
A Survey of Inductive Reasoning for Large Language Models
Survey of inductive reasoning in LLMs, covering particular-to-general thinking patterns and knowledge generalization capabilities.
4/14/2026
Domain-Specific Data Generation Framework for RAG Adaptation
RAGen framework for generating domain-specific question-answer pairs to adapt RAG systems to specialized applications.
4/14/2026
Learning When Not to Learn: Risk-Sensitive Abstention in Bandits with Unbounded Rewards
Risk-sensitive abstention in bandit algorithms for high-stakes AI where errors are irreparable without expert guidance.
4/14/2026
Think Parallax: Solving Multi-Hop Problems via Multi-View Knowledge-Graph-Based Retrieval-Augmented Generation
Multi-hop reasoning over knowledge graphs using multi-view RAG with LLMs, addressing Transformer attention specialization patterns.
4/14/2026
SimBench: Benchmarking the Ability of Large Language Models to Simulate Human Behaviors
SimBench provides first standardized benchmark for evaluating how faithfully LLMs simulate human behaviors across diverse tasks and metrics.
4/14/2026
AtlasKV: Augmenting LLMs with Billion-Scale Knowledge Graphs in 20GB VRAM
AtlasKV enables RAG systems to integrate billion-scale knowledge graphs efficiently in limited VRAM by avoiding expensive external retrieval modules.
4/14/2026
What's In My Human Feedback? Learning Interpretable Descriptions of Preference Data
Method to automatically extract and explain what features human feedback data encodes when training language models, addressing unpredictability in RLHF approaches.