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。
Seungone Kim, Juyoung Suk, Shayne Longpre +7 authors
Prometheus 2 is an advanced open-source evaluator LM that aligns closely with human and proprietary LM judgments, supporting both direct assessment and pairwise ranking with customizable criteria.
43 篇论文 · 按点赞排序
Justin Zhao, Timothy Wang, Wael Abid +7 authors
LoRA fine-tuning reduces computational resources while maintaining high performance across various tasks and models, and LoRAX allows for efficient deployment of multiple LLMs on a single GPU.
Wei Chen, Zhiyuan Li
The Octopus v4 model uses functional tokens to integrate and direct queries to task-specific open-source language models, achieving SOTA performance with models under 10B parameters.
Ziming Liu, Yixuan Wang, Sachin Vaidya +5 authors
Kolmogorov-Arnold Networks (KANs) outperform Multi-Layer Perceptrons (MLPs) in accuracy and interpretability by using learnable activation functions and spline-based weights.
Fabian Gloeckle, Badr Youbi Idrissi, Baptiste Rozière +2 authors
Training large language models to predict multiple future tokens simultaneously enhances their sample efficiency and performance, particularly on generative benchmarks and small algorithmic tasks, with decreased inference time.
Chanran Kim, Jeongin Lee, Shichang Joung +2 authors
InstantFamily uses masked cross-attention and multimodal embeddings to generate images with preserved multi-IDs and high visual coherence.
Pat Verga, Sebastian Hofstatter, Sophia Althammer +6 authors
Evaluating Large Language Models using a Panel of smaller evaluators outperforms single large evaluators with reduced bias and cost.
Wenting Zhao, Xiang Ren, Jack Hessel +3 authors
Chatbots such as GPT-4 and ChatGPT are now serving millions of users. Despite their widespread use, there remains a lack of public datasets showcasing how these tools are used by a population of users in practice. To bridge this gap, we offered free access to ChatGPT for online users in exchange for their affirmative, consensual opt-in to anonymously collect their chat transcripts and request headers. From this, we compiled WildChat, a corpus of 1 million user-ChatGPT conversations, which consists of over 2.5 million interaction turns. We compare WildChat with other popular user-chatbot interaction datasets, and find that our dataset offers the most diverse user prompts, contains the largest number of languages, and presents the richest variety of potentially toxic use-cases for researchers to study. In addition to timestamped chat transcripts, we enrich the dataset with demographic data, including state, country, and hashed IP addresses, alongside request headers. This augmentation allows for more detailed analysis of user behaviors across different geographical regions and temporal dimensions. Finally, because it captures a broad range of use cases, we demonstrate the dataset's potential utility in fine-tuning instruction-following models. WildChat is released at https://wildchat.allen.ai under AI2 ImpACT Licenses.
Yupeng Zhou, Daquan Zhou, Ming-Ming Cheng +2 authors
StoryDiffusion, combining consistent self-attention and semantic motion prediction, enables generation of coherent and stable images and videos from textual descriptions.
Richard Yuanzhe Pang, Weizhe Yuan, Kyunghyun Cho +3 authors
An iterative preference optimization method using a modified DPO loss improves reasoning accuracy on various datasets by optimizing winning and losing reasoning steps in Chain-of-Thought candidates.
Lin Xu, Yilin Zhao, Daquan Zhou +3 authors
Pooling LLaVA improves video-language pre-trained models by smoothing high-norm visual features, achieving state-of-the-art performance in video question-answering, captioning, and benchmark tasks.
Peitian Zhang, Ninglu Shao, Zheng Liu +4 authors
Llama-3-8B-Instruct's context length is extended from 8K to 80K using QLoRA fine-tuning with minimal additional training samples, demonstrating significant potential for further context extension with increased computational resources.
Hugh Zhang, Jeff Da, Dean Lee +12 authors
Evaluation of large language models on a new math benchmark, GSM1k, reveals that many models exhibit signs of overfitting to the existing GSM8k benchmark, leading to performance drops on GSM1k.
Sheng-Chieh Lin, Luyu Gao, Barlas Oguz +4 authors
Proposes factuality-aware alignment methods for pre-trained language models to enhance factual accuracy during supervised fine-tuning and reinforcement learning, reducing hallucination.
Fangcheng Liu, Yehui Tang, Zhenhua Liu +3 authors
Kangaroo, a self-speculative decoding framework using a shallow sub-network as a self-draft model, accelerates large language model inference with improved token acceptance and reduced additional parameters.
Gerald Shen, Zhilin Wang, Olivier Delalleau +10 authors
NeMo-Aligner is a toolkit for aligning large language models with human values using scalable techniques like RLHF, DPO, SteerLM, and SPIN, optimized for use with hundreds of GPUs and parameter-efficient fine-tuning.
Yue Wu, Zhiqing Sun, Huizhuo Yuan +3 authors
A self-play method called SPPO for language model alignment achieves state-of-the-art performance by approximating Nash equilibrium policy in a constant-sum game setting, outperforming other approaches with limited data.
Anselm Paulus, Arman Zharmagambetov, Chuan Guo +2 authors
A novel method uses AdvPrompter, another LLM, to generate human-readable adversarial prompts efficiently, improving the state-of-the-art in generating unpredictable yet harmful outputs, and enhancing LLM robustness against jailbreaking attacks.
Navve Wasserman, Noam Rotstein, Roy Ganz +1 authors
A diffusion model is trained to add objects to images based on text descriptions using a large-scale dataset of images with removed objects, outperforming existing methods.
Wenxun Dai, Ling-Hao Chen, Jingbo Wang +3 authors
MotionLCM generates real-time controllable human motions using text and control signals by extending latent diffusion models with a Motion ControlNet.
Khaled Saab, Tao Tu, Wei-Hung Weng +63 authors
Med-Gemini, a family of specialized multimodal AI models, achieves state-of-the-art performance across various medical benchmarks, surpassing GPT-4 and demonstrating superior long-context and uncertainty-guided search capabilities.
Yunhao Ge, Xiaohui Zeng, Jacob Samuel Huffman +3 authors
VisualFactChecker improves caption accuracy and detail through a three-step pipeline combining image-to-text models, fact-checking with large language models, and final summarization, outperforming existing methods on multiple datasets.
Zhili Cheng, Zhitong Wang, Jinyi Hu +7 authors
LEGENT is an open platform for developing embodied agents using LLMs and LMMs, featuring a rich 3D environment and a data generation pipeline that outperforms GPT-4V in embodied tasks.
Maxwell Jones, Sheng-Yu Wang, Nupur Kumari +2 authors
A new method called Pair Customization learns and applies stylistic differences from a single image pair to a generative model without overfitting to specific image content.
Junsang Yoon, Akshat Gupta, Gopala Anumanchipalli
The analysis of ROME, MEMIT, and EMMET editing techniques on Llama-3 highlights that sequential editing preserves model performance better than large batch editing across various strategies.
Kai Zhang, Sai Bi, Hao Tan +4 authors
GS-LRM is a scalable transformer-based model that predicts high-quality 3D Gaussian primitives from sparse images, handling both objects and scenes effectively.
Peng Chu, Jiang Wang, Andre Abrantes
An automated pipeline using GPT-4V and tracking-based character recognition generates high-quality, contextually consistent audio descriptions without additional training.
Ian Huang, Guandao Yang, Leonidas Guibas
A system using Vision-Language Models and image-generation models automates Blender scene editing by finding the correct sequence of actions to achieve user-defined goals.
Haohe Liu, Xuenan Xu, Yi Yuan +3 authors
SemantiCodec is a dual-encoder codec that compresses diverse audio types into fewer tokens using AudioMAE and k-means clustering, improved by a diffusion-model-based decoder, achieving higher semantic information and superior reconstruction quality at lower bitrates than existing codecs.
Bin Xiao, Chunan Shi, Xiaonan Nie +5 authors
Clover, a new speculative decoding algorithm, enhances the efficiency of large language models by integrating sequential knowledge into parallel decoding, improving hit rates and outperforming previous methods.
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