A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Jos'ephine Raugel, Marc Szafraniec, Huy V. Vo, et al.
This paper systematically varies model size, training amount, and image type in DINOv3 vision transformers to disentangle how these factors independently and interactively drive brain-like representations, revealing a developmental chronology aligned
Narges Ahmadi, Yubo Jiao, Jônatas Augusto Manzolli, et al.
This paper proposes a three-agent workflow integrating conversational data collection, structured processing, and LLM-based prediction for travel behavior, achieving 71.5% accuracy with vision-based LLMs.
Adnan El Assadi, Niklas Muennighoff, Jinhyuk Lee
A controlled, cost-aware comparison of LLMs and embedding models across 37 tasks finds them effectively tied in quality, but LLMs cost up to 1,431x more, supporting a division of labor.
Hmrishav Bandyopadhyay, Xuanchi Ren, Zijian Huang, et al.
Introduces Context-Matched Distillation (CMD), a causal DMD framework that aligns teacher supervision with the information available during autoregressive video generation, improving control adherence and state-of-the-art performance.
Yan Deng, Fei Xu
DreamFly introduces causal memory and receding-horizon diffusion planning to improve aerial vision-language navigation, achieving state-of-the-art results on the OpenFly benchmark.
Jy-oti Aneja, Michael Harrison, Neel Joshi, et al.
Phi-4-reasoning-vision-15B is a compact open-weight multimodal reasoning model demonstrating that careful architecture and data curation enable competitive performance with significantly less compute.
Yijiang Li, Bingyang Wang, Yijun Liang, et al.
Proposes U-OPSD, an unsupervised on-policy self-distillation method that uses only a model's own generations and majority voting to improve reasoning without external supervision.
Juan S. Santillana
VectraYX-Vision-1B is a sub-2B Spanish/LATAM cybersecurity VLM with structured reasoning and tool use, but current vision SFT yields near-zero visual grounding.
Unknown
Data2vec is a general self-supervised learning framework that uses the same learning method for speech, NLP, and computer vision.
Unknown
This paper presents a comprehensive recipe for applying sim-to-real reinforcement learning to vision-based dexterous manipulation on humanoid robots, synthesizing practical techniques and insights.
Unknown
This paper introduces a sim-to-real approach using randomized-to-canonical adaptation networks to transfer a vision-based closed-loop grasping reinforcement learning agent from simulation to a real robot.
Unknown
This paper introduces Fast, an efficient action tokenization method for vision-language-action models, aiming to improve the speed and practicality of generalist robot policies.