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。
403 篇论文 · 按点赞排序
Subin Kim, Kyungmin Lee, June Suk Choi +3 authors
A novel method, Collaborative Score Distillation (CSD), based on Stein Variational Gradient Descent (SVGD), enhances consistency in text-to-image diffusion models across multiple images such as panoramas, videos, and 3D scenes.
Scott L. Fleming, Alejandro Lozano, William J. Haberkorn +27 authors
MedAlign, a curated dataset of natural language instructions for EHR data, evaluates the accuracy and quality of general domain LLMs in healthcare, revealing high error rates and context length dependencies.
Yang Liu, Yuanshun Yao, Jean-Francois Ton +5 authors
This survey examines key dimensions of trustworthiness for large language models, including reliability, safety, fairness, misuse resistance, explainability, social norms adherence, and robustness, providing guidelines for evaluating and improving model alignment.
Aliasghar Khani, Saeid Asgari Taghanaki, Aditya Sanghi +2 authors
SLiMe segments images at desired granularity using Stable Diffusion with minimal annotations and outperforms existing one-shot and few-shot segmentation methods.
Bill Yuchen Lin, Abhilasha Ravichander, Ximing Lu +5 authors
A study explores the effectiveness of alignment tuning in large language models, suggesting that superficial methods like in-context learning can achieve strong performance without traditional supervised fine-tuning or reinforcement learning.
Jiasen Lu, Christopher Clark, Sangho Lee +5 authors
A unified autoregressive multimodal model handles image, text, audio, and action by tokenizing inputs into a shared space and processing them with a transformer architecture, achieving state-of-the-art performance across various benchmarks.
Katherine Tian, Eric Mitchell, Huaxiu Yao +2 authors
Fine-tuning language models using automatically generated factuality preference rankings improves their factual accuracy without human labeling.
Sang-Hoon Lee, Ha-Yeong Choi, Seung-Bin Kim +1 authors
HierSpeech++, an enhanced hierarchical variational autoencoder, achieves high-quality zero-shot speech synthesis by combining self-supervised representations and an efficient super-resolution framework.
Ke Wang, Houxing Ren, Aojun Zhou +7 authors
A method is presented for fine-tuning language models to solve math problems using code, achieving state-of-the-art results on MATH and GSM8K datasets.
Yixiao Li, Yifan Yu, Chen Liang +4 authors
LoftQ, a novel quantization framework, improves performance of LLMs by simultaneously quantizing and initializing LoRA fine-tuning, reducing the gap in downstream task performance.
Weijia Shi, Sewon Min, Maria Lomeli +7 authors
In-Context Pretraining improves large language models by ordering documents to maximize contextual similarity, leading to better performance in complex reasoning tasks.
Dawid Jan Kopiczko, Tijmen Blankevoort, Yuki Markus Asano
Vector-based Random Matrix Adaptation (VeRA) reduces the number of trainable parameters by 10x compared to LoRA while maintaining performance, and is demonstrated on benchmarks like GLUE and E2E, showing its utility in instruction-following.
Rui Zheng, Shihan Dou, Songyang Gao +24 authors
This report examines Reinfocement Learning with Human Feedback (RLHF) and proposes PPO-max to improve the stability of policy model training compared to other SFT models and ChatGPT.
Antoine Guédon, Vincent Lepetit
A method for fast and precise mesh extraction from 3D Gaussian Splatting using a regularization term and Poisson reconstruction, enabling real-time editing and rendering quality.
Risto Luukkonen, Ville Komulainen, Jouni Luoma +18 authors
The study addresses the challenges of creating large language models for underrepresented languages like Finnish, developing both monolingual and multilingual models, and evaluating their performance through a newly created benchmark.
Yufei Wang, Zhou Xian, Feng Chen +6 authors
RoboGen automates the learning of diverse robotic skills at scale using generative simulation to propose tasks and generate training environments and supervision.
Ying Sheng, Shiyi Cao, Dacheng Li +9 authors
S-LoRA is a system that allows for efficient and scalable serving of numerous LoRA adapters using a unified memory pool, tensor parallelism, and custom CUDA kernels.
Da Yin, Faeze Brahman, Abhilasha Ravichander +4 authors
Lumos, a modular language agent framework using open-source LLMs, achieves superior performance across various tasks and demonstrates enhanced generalization.
Hao Ouyang, Qiuyu Wang, Yuxi Xiao +6 authors
CoDeF represents videos using a canonical content field and a temporal deformation field, enabling the lifting of image processing algorithms to video processing with superior cross-frame consistency.
Neha Sengupta, Sunil Kumar Sahu, Bokang Jia +19 authors
Jais and Jais-chat, Arabic-centric LLMs based on the GPT-3 architecture, outperform existing models in Arabic and are competitive in English with fewer resources.
Michaël Mathieu, Sherjil Ozair, Srivatsan Srinivasan +21 authors
A benchmark called AlphaStar Unplugged is established for offline reinforcement learning using a dataset from StarCraft II, featuring unprecedented challenges and achieving high win rates with offline data.
Tamera Lanham, Anna Chen, Ansh Radhakrishnan +27 authors
Larger language models may not produce faithful reasoning when using chain-of-thought despite performance improvements, depending on model size and task.
Yang Zhao, Zhijie Lin, Daquan Zhou +3 authors
BuboGPT is a multi-modal LLM with visual grounding that enhances understanding and interaction between vision, audio, and language.
Tao Ge, Jing Hu, Xun Wang +2 authors
The In-context Autoencoder (ICAE) compresses long contexts into memory slots for efficient LLM inference using a combination of pretraining and fine-tuning.
Guillaume Sanchez, Honglu Fan, Alexander Spangher +3 authors
Classifier-Free Guidance enhances performance across various language modeling tasks and improves the faithfulness and coherence of AI assistants, outperforming models with higher parameter counts.
Ho Kei Cheng, Seoung Wug Oh, Brian Price +2 authors
DEVA, a decoupled approach using image-level segmentation and bi-directional temporal propagation, achieves favorable results in various data-scarce video segmentation tasks with reduced annotation and training costs.
Zhongfei Qing, Zhongang Cai, Zhitao Yang +1 authors
A new system combines language models and motion retrieval to generate long, stable, and semantically aligned human motions from text descriptions, outperforming previous methods in trajectory, action, and blending tasks.
Chenxu Zhang, Chao Wang, Jianfeng Zhang +7 authors
DREAM-Talk is a two-stage diffusion-based framework generating expressive emotional talking faces with accurate lip-sync from single portrait images.
Avi Singh, John D. Co-Reyes, Rishabh Agarwal +35 authors
Self-training with scalar feedback using expectation-maximization improves language model performance on tasks with verifiable answers, reducing reliance on human data.
William Berrios, Gautam Mittal, Tristan Thrush +2 authors
LENS uses language models to reason over outputs from vision modules, achieving competitive performance in vision and vision-language tasks without multimodal training.
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