This paper introduces EUGens: Efficient, Unified and General Dense Layers, capable of accurately approximating general feedforward fully connected layers, but of linear rather than quadratic time complexity.

We introduce 4D-LRM, the first large-scale 4D reconstruction model that takes input from unconstrained views and timestamps and renders arbitrary novel view-time combinations.

We introduce MJ-Bench-Video, a large-scale video preference dataset for comprehensively evaluating the reward models of text-to-video generation as well as a MoE-based video reward model

A novel benchmark using a comprehensive preference dataset to evaluate multimodal judges across multiple key perspectives

We propose a diffusion-based topology-aware graph generation method that aims to closely resemble the structural characteristics of the original graph by leveraging persistent homology from topological data analysis (TDA).

This paper utilizes multi-agent debate process for llm-as-judge, and employs an adaptive stopping mechanism.

ReAgent-V enables reward-driven, multi-agent video understanding with dynamic reflection and frame selection.

NFL-BA enhances tracking and mapping in neural rendering-based SLAM by explicitly modeling dynamic near-field lighting, with applications in endoscopy.

This paper introduces a framework to analyze the trustworthiness of near-optimal sparse decision trees from the Rashomon set, showing they can perform as well as models explicitly optimized for fairness, robustness, or privacy.