AI研究論文デイリー

Data Advisor: Dynamic Data Curation for Safety Alignment of Large Language Models

DART: Denoising Autoregressive Transformer for Scalable Text-to-Image Generation

Vector-ICL: In-context Learning with Continuous Vector Representations

Benchmarking Agentic Workflow Generation

Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition

DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models

Preserving Multi-Modal Capabilities of Pre-trained VLMs for Improving Vision-Linguistic Compositionality

Optima: Optimizing Effectiveness and Efficiency for LLM-Based Multi-Agent System

Rectified Diffusion: Straightness Is Not Your Need in Rectified Flow

LPZero: Language Model Zero-cost Proxy Search from Zero

MLLM as Retriever: Interactively Learning Multimodal Retrieval for Embodied Agents

MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical Code

Progressive Autoregressive Video Diffusion Models

Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates

Zebra: In-Context and Generative Pretraining for Solving Parametric PDEs

Intriguing Properties of Large Language and Vision Models

Agent S: An Open Agentic Framework that Uses Computers Like a Human

Emergent properties with repeated examples

Towards Self-Improvement of LLMs via MCTS: Leveraging Stepwise Knowledge with Curriculum Preference Learning

PrefixQuant: Static Quantization Beats Dynamic through Prefixed Outliers in LLMs

WALL-E: World Alignment by Rule Learning Improves World Model-based LLM Agents

SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe

Scaling Up Your Kernels: Large Kernel Design in ConvNets towards Universal Representations

Accelerated Preference Optimization for Large Language Model Alignment

GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models

MotionGS: Exploring Explicit Motion Guidance for Deformable 3D Gaussian Splatting