AI研究論文デイリー

Video SimpleQA: Towards Factuality Evaluation in Large Video Language Models

Aether: Geometric-Aware Unified World Modeling

Variance Control via Weight Rescaling in LLM Pre-training

Position: Interactive Generative Video as Next-Generation Game Engine

Mind with Eyes: from Language Reasoning to Multimodal Reasoning

DynamicVis: An Efficient and General Visual Foundation Model for Remote Sensing Image Understanding

FFN Fusion: Rethinking Sequential Computation in Large Language Models

AgentRxiv: Towards Collaborative Autonomous Research

Equivariant Image Modeling

OmnimatteZero: Training-free Real-time Omnimatte with Pre-trained Video Diffusion Models

Judge Anything: MLLM as a Judge Across Any Modality

RDTF: Resource-efficient Dual-mask Training Framework for Multi-frame Animated Sticker Generation

Training-free Diffusion Acceleration with Bottleneck Sampling

AMD-Hummingbird: Towards an Efficient Text-to-Video Model

Instruct-CLIP: Improving Instruction-Guided Image Editing with Automated Data Refinement Using Contrastive Learning

Revisiting Image Fusion for Multi-Illuminant White-Balance Correction

Defeating Prompt Injections by Design

Reasoning to Learn from Latent Thoughts

Optimized Minimal 3D Gaussian Splatting

Diffusion-4K: Ultra-High-Resolution Image Synthesis with Latent Diffusion Models

Feather-SQL: A Lightweight NL2SQL Framework with Dual-Model Collaboration Paradigm for Small Language Models

Typed-RAG: Type-aware Multi-Aspect Decomposition for Non-Factoid Question Answering

I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders

Verbal Process Supervision Elicits Better Coding Agents

QuartDepth: Post-Training Quantization for Real-Time Depth Estimation on the Edge

Lost in Cultural Translation: Do LLMs Struggle with Math Across Cultural Contexts?

CFG-Zero*: Improved Classifier-Free Guidance for Flow Matching Models

SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild

V-Seek: Accelerating LLM Reasoning on Open-hardware Server-class RISC-V Platforms

Video-T1: Test-Time Scaling for Video Generation

LEMMA: Learning from Errors for MatheMatical Advancement in LLMs

Human Motion Unlearning

Vision-R1: Evolving Human-Free Alignment in Large Vision-Language Models via Vision-Guided Reinforcement Learning

AlphaSpace: Enabling Robotic Actions through Semantic Tokenization and Symbolic Reasoning

Rethinking Image Evaluation in Super-Resolution

MagicComp: Training-free Dual-Phase Refinement for Compositional Video Generation

CODA: Repurposing Continuous VAEs for Discrete Tokenization

Global-Local Tree Search for Language Guided 3D Scene Generation

MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse