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。
Dirk Groeneveld, Iz Beltagy, Pete Walsh +40 authors
OLMo, an open-sourced language model, provides comprehensive access to training data, code, and architecture, facilitating research and innovation in language modeling.
50 篇论文 · 按点赞排序
Saleh Ashkboos, Maximilian L. Croci, Marcelo Gennari do Nascimento +2 authors
SliceGPT, a novel post-training sparsification method, reduces model parameters by up to 25% without significant performance decrease, optimizing inference speed and GPU usage on large language models.
Luca Soldaini, Rodney Kinney, Akshita Bhagia +33 authors
Dolma, a three trillion token English corpus, and its open-source data curation toolkit facilitate open research in language model pretraining by providing detailed insights into data curation practices.
Xiaoyi Dong, Pan Zhang, Yuhang Zang +20 authors
InternLM-XComposer2, a vision-language model, excels in free-form text-image creation and understanding using Partial LoRA, surpassing existing models and matching GPT-4V and Gemini Pro in multimodal benchmarks.
Bin Lin, Zhenyu Tang, Yang Ye +6 authors
MoE-LLaVA, a sparse Large Vision-Language Model architecture, improves performance and reduces hallucinations with fewer active parameters compared to larger models.
Pratyush Maini, Skyler Seto, He Bai +3 authors
Web Rephrase Augmented Pre-training (WRAP) improves large language model pre-training efficiency and performance by incorporating synthetic rephrases styled like Wikipedia or question-answer format into the training dataset.
Parth Sarthi, Salman Abdullah, Aditi Tuli +3 authors
The RAPTOR model enhances retrieval-augmented language models by recursively embedding, clustering, and summarizing text chunks, leading to better performance on question-answering tasks involving complex reasoning.
Tiannan Wang, Jiamin Chen, Qingrui Jia +43 authors
Weaver, a family of specialized large language models, achieves superior writing capabilities through pre-training and fine-tuning methods, surpasses GPT-4 in various writing tasks, and supports retrieval-augmented generation and tool usage.
Tianheng Cheng, Lin Song, Yixiao Ge +3 authors
YOLO-World enhances YOLO with open-vocabulary detection using vision-language modeling and a re-parameterizable network, achieving high accuracy and speed.
Xiaoyu Shi, Zhaoyang Huang, Fu-Yun Wang +9 authors
Motion-I2V, a diffusion-based framework for image-to-video generation, achieves consistency and controllability through explicit motion modeling and trajectory prediction.
Jiacheng Liu, Sewon Min, Luke Zettlemoyer +2 authors
Modernized n-gram models, specifically $\infty$-gram models, enhance text analysis and neural LLMs by providing high accuracy in next-token prediction and reducing language modeling perplexities.
Chaochao Lu, Chen Qian, Guodong Zheng +33 authors
A qualitative study evaluates the generalizability, trustworthiness, and causal reasoning capabilities of MLLMs across text, code, image, and video modalities to understand their reliability for downstream applications.
Zhennan Wu, Yang Li, Han Yan +8 authors
BlockFusion, a diffusion-based model, generates and expands 3D scenes using latent tri-plane representations and a 2D layout conditioning mechanism.
Manuel Faysse, Patrick Fernandes, Nuno Guerreiro +13 authors
CroissantLLM, a 1.3B bilingual language model trained on balanced English and French datasets, achieves high performance on consumer hardware and is validated using custom transparency criteria in FrenchBench.
Tairan He, Chong Zhang, Wenli Xiao +3 authors
A learning-based control framework, Agile But Safe (ABS), enables quadrupedal robots to navigate cluttered environments safely and at high speeds by integrating agile and recovery policies guided by a reach-avoid value network.
Jianlan Luo, Zheyuan Hu, Charles Xu +7 authors
A library for robotic reinforcement learning provides efficient off-policy deep RL methods, high-quality controllers, and example tasks, achieving state-of-the-art results in tasks like PCB board assembly and cable routing.
Yushi Bai, Xin Lv, Jiajie Zhang +6 authors
LongAlign improves the handling of long contexts in large language models through specialized instruction datasets, packing and batching strategies, loss weighting, and a benchmark for extended queries.
Yilun Zhu, Joel Ruben Antony Moniz, Shruti Bhargava +6 authors
The benchmark evaluates LLMs' context understanding by assessing pre-trained and quantized models across four tasks and nine datasets, highlighting the performance differences between dense and fine-tuned models.
Qingcheng Zhao, Pengyu Long, Qixuan Zhang +6 authors
The proposed Generalized Neural Parametric Facial Asset and Media2Face diffusion model generate high-fidelity and expressive 3D facial animations from speech with rich multi-modality guidance.
Vikranth Dwaracherla, Seyed Mohammad Asghari, Botao Hao +1 authors
Efficient exploration using double Thompson sampling and epistemic neural network uncertainty estimation significantly improves large language models with reduced query numbers.
Fu-Yun Wang, Zhaoyang Huang, Xiaoyu Shi +4 authors
AnimateLCM, an extension of Latent Consistency Model, achieves high-fidelity video generation efficiently with decoupled consistency learning and adaptability to existing adapters, outperforming in image-conditioned and layout-conditioned video generation.
Jordi Grau-Moya, Tim Genewein, Marcus Hutter +8 authors
Meta-learning can leverage Universal Turing Machine-generated data to train neural networks for universal prediction strategies.
Marius-Constantin Dinu, Claudiu Leoveanu-Condrei, Markus Holzleitner +2 authors
SymbolicAI integrates generative models with solvers using LLMs as semantic parsers, enabling symbolic reasoning and generative AI tasks via probabilistic and differentiable programming.
Silin Gao, Jane Dwivedi-Yu, Ping Yu +7 authors
Chain-of-Abstraction (CoA) method enhances large language models' multi-step reasoning by using abstract reasoning chains to plan and efficiently invoke tools, improving accuracy and speed in tasks like mathematical reasoning and Wiki QA.
Zecheng Tang, Chenfei Wu, Zekai Zhang +8 authors
StrokeNUWA introduces stroke tokens for vector graphic representation, enhancing semantic coherence in LLM-based visual synthesis and achieving significant speed and compression improvements.
Junyang Wang, Haiyang Xu, Jiabo Ye +5 authors
Mobile-Agent, a vision-centric multimodal large language model agent for mobile devices, autonomously performs complex tasks without relying on system-specific customizations and achieves high accuracy and adaptability.
Yuhui Li, Fangyun Wei, Chao Zhang +1 authors
EAGLE framework accelerates large language model inference by operating at a feature level with lossless performance relative to vanilla auto-regressive decoding.
Xiaoyu Li, Qi Zhang, Di Kang +7 authors
A survey of fundamental methodologies in 3D generation, including 3D representations, generation paradigms, datasets, applications, and challenges.
Robin San Roman, Pierre Fernandez, Alexandre Défossez +3 authors
AudioSeal is an audio watermarking technique that uses a generator/detector architecture to detect AI-generated speech with high robustness and imperceptibility, offering fast detection.
Philipp Singer, Pascal Pfeiffer, Yauhen Babakhin +4 authors
A 1.8B parameter language model trained on 1T tokens demonstrates competitive performance across benchmarks and is released publicly with supervised fine-tuning and preference optimization.
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