Explore · 每周精选
发现最受关注的研究论文,追踪研究趋势,订阅感兴趣的期刊与关键词。
Lihe Yang, Bingyi Kang, Zilong Huang +4 authors
Depth Anything V2 improves monocular depth estimation through synthetic images, larger teacher models, and pseudo-labeled real images, achieving better efficiency and accuracy than Stable Diffusion models.
50 篇论文 · 按点赞排序
Zhangchen Xu, Fengqing Jiang, Luyao Niu +4 authors
Magpie, a self-synthesis method, generates high-quality instruction data from aligned LLMs, enabling competitive performance in alignment benchmarks compared to models fine-tuned with larger, supervised datasets.
Peize Sun, Yi Jiang, Shoufa Chen +4 authors
LlamaGen applies the next-token prediction paradigm from large language models to image generation, achieving state-of-the-art performance with various model sizes and conditions.
Sander Schulhoff, Michael Ilie, Nishant Balepur +28 authors
A taxonomy and meta-analysis are provided for prompting techniques and vocabulary in GenAI systems, with a focus on natural language prefix-prompting.
Qihang Yu, Mark Weber, Xueqing Deng +3 authors
Transformer-based 1D Tokenizer (TiTok) reduces computational demands and enhances image synthesis efficiency by significantly decreasing token count and outperforming state-of-the-art methods in both speed and quality.
Junlin Wang, Jue Wang, Ben Athiwaratkun +2 authors
A Mixture-of-Agents (MoA) methodology combines multiple large language models (LLMs) to achieve state-of-the-art performance on various evaluations.
Duy-Kien Nguyen, Mahmoud Assran, Unnat Jain +3 authors
Vanilla Transformers achieve high performance in computer vision tasks by treating individual pixels as tokens, challenging the conventional use of local neighborhoods.
Ting-Hsuan Chen, Jiewen Chan, Hau-Shiang Shiu +3 authors
A video editing framework combining deformation fields and diffusion priors produces high-quality canonical images for diverse video editing tasks with accelerated training using low-rank adaptation.
Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti +15 authors
OpenVLA, a 7B-parameter open-source vision-language-action model, demonstrates strong performance in generalist manipulation and efficient fine-tuning for new tasks, outperforming larger closed models and from-scratch imitation learning methods.
Xiao Yang, Kai Sun, Hao Xin +24 authors
The Comprehensive RAG Benchmark (CRAG) evaluates RAG solutions with diverse QA tasks, revealing gaps in LLM accuracy and future research directions.
Wilfried Bounsi, Borja Ibarz, Andrew Dudzik +5 authors
A novel TransNAR model combines Transformer-based language understanding with graph neural network solvers to enhance algorithmic reasoning.
Xianhang Li, Haoqin Tu, Mude Hui +9 authors
Enhancing image-text pairs through recaptioning with LLaMA-3 improves model performance in vision-language tasks, including zero-shot retrieval and text-to-image generation.
Pengyang Ling, Jiazi Bu, Pan Zhang +6 authors
MotionClone, a training-free framework, uses motion cloning to control text-to-video generation, yielding superior motion fidelity, textual alignment, and temporal consistency.
Liliang Ren, Yang Liu, Yadong Lu +3 authors
Samba, a hybrid architecture combining State Space Models and Sliding Window Attention, efficiently handles long-range context with lower computational complexity and better memory recall than pure attention or SSM models.
Zhenliang Xue, Yixin Song, Zeyu Mi +3 authors
PowerInfer-2 is a framework that enhances the high-speed inference of large language models on smartphones by utilizing heterogeneous resources and optimizing matrix computations through neuron cluster processing.
Fangfu Liu, Hanyang Wang, Shunyu Yao +3 authors
Physics3D uses a video diffusion model to learn and simulate the physical properties of 3D objects, enabling realistic behavior prediction in virtual environments.
Linzheng Chai, Shukai Liu, Jian Yang +15 authors
A multilingual code benchmark covering 40 programming languages with 16K test samples is introduced to advance code language model research, along with a multilingual coder model and instruction corpora.
Aryo Pradipta Gema, Joshua Ong Jun Leang, Giwon Hong +13 authors
Analysis identifies numerous errors in the MMLU benchmark, leading to the creation of MMLU-Redux, a re-annotated subset that highlights discrepancies in model performance metrics.
Zesen Cheng, Sicong Leng, Hang Zhang +8 authors
VideoLLaMA 2, an enhanced Video Large Language Model, incorporates Spatial-Temporal Convolution and audio cues to achieve competitive performance in multimodal video and audio tasks.
Xi Chen, Yutong Feng, Mengting Chen +5 authors
MimicBrush, a generative training framework developed from a diffusion prior, enables imitative image editing by leveraging video frames and masked regions to capture semantic correspondence and improve editing quality.
Mert Yuksekgonul, Federico Bianchi, Joseph Boen +4 authors
TextGrad, a novel framework, uses text-based feedback from LLMs to optimize variables in compound AI systems across diverse applications, enhancing performance and generality.
Qihao Liu, Zhanpeng Zeng, Ju He +3 authors
A novel multi-resolution network and time-dependent layer normalization enhance diffusion models, improving image generation fidelity and efficiency.
Jianing Yang, Xuweiyi Chen, Nikhil Madaan +4 authors
3D-GRAND, a large-scale dataset, enhances 3D-LLM performance and hallucination reduction through instruction tuning and establishes a benchmark for sim-to-real transfer.
Di Zhang, Jiatong Li, Xiaoshui Huang +3 authors
MCT Self-Refine (MCTSr) combines Large Language Models with Monte Carlo Tree Search to enhance accuracy and reliability in complex mathematical reasoning by leveraging systematic exploration and heuristic self-refinement.
Bill Yuchen Lin, Yuntian Deng, Khyathi Chandu +6 authors
WildBench evaluates large language models using human-chatbot queries, providing tasks and metrics that correlate well with human-voted evaluations.
Bahare Fatemi, Mehran Kazemi, Anton Tsitsulin +6 authors
Synthetic datasets and an evaluation framework are introduced to assess Large Language Models' capabilities in various temporal reasoning scenarios.
Liangchen Luo, Yinxiao Liu, Rosanne Liu +8 authors
A novel divide-and-conquer Monte Carlo Tree Search algorithm called OmegaPRM enables efficient collection of process supervision data to enhance the reasoning performance of LLMs without human intervention.
Yixin Song, Haotong Xie, Zhengyan Zhang +4 authors
A novel activation function (dReLU) and sparsification method improve the efficiency and speed of large language models without sacrificing performance.
Xuehai He, Weixi Feng, Kaizhi Zheng +11 authors
MMWorld is a new benchmark for evaluating multimodal video understanding in various disciplines and reasoning tasks, highlighting gaps between current models and human performance.
Joongwon Kim, Bhargavi Paranjape, Tushar Khot +1 authors
Husky, an open-source language agent, excels in complex reasoning tasks across multiple domains by iteratively generating and executing actions, outperforming existing models and matching GPT-4 on a comprehensive evaluation set.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号