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。
50 篇论文 · 按点赞排序
Qianli Ma, Yaowei Zheng, Zhelun Shi +9 authors
A modular training framework accelerates the development of omni-modal LLMs through efficient 3D parallelism and flexible configuration.
Yuhang Ma, Xiaoshi Wu, Keqiang Sun +1 authors
HPSv3, a human preference score using a wide-spectrum dataset and uncertainty-aware ranking loss, enhances text-to-image generation quality through iterative refinement.
Linan Yue, Yichao Du, Yizhi Wang +8 authors
Research on efficient reasoning methods for Large Reasoning Models (LRMs) aims to reduce reasoning path length without sacrificing performance, through single-model optimization and model collaboration.
Yuhan Guo, Cong Guo, Aiwen Sun +12 authors
A framework for web agents decomposes their capabilities into knowledge content learning and cognitive processes, using a structured dataset and a novel reasoning framework to enhance generalization and performance.
Guozhao Mo, Wenliang Zhong, Jiawei Chen +6 authors
LiveMCPBench provides a comprehensive benchmark for evaluating LLM agents across a diverse set of real-world tasks in the MCP ecosystem, using a scalable evaluation pipeline and adaptive judging framework.
Zexiong Ma, Chao Peng, Qunhong Zeng +3 authors
ToolTrain, a two-stage training framework combining supervised fine-tuning and reinforcement learning, enhances LLMs for issue localization by integrating repository retrieval tools, achieving state-of-the-art performance.
Maziyar Panahi
OpenMed NER, a suite of open-source transformer models using DAPT and LoRA, achieves state-of-the-art performance on diverse biomedical NER benchmarks with high efficiency and low computational cost.
Bowen Zhang, Sicheng Xu, Chuxin Wang +4 authors
A novel framework uses a Direct 4DMesh-to-GS Variation Field VAE and Gaussian Variation Field diffusion model to generate high-quality dynamic 3D content from single video inputs, demonstrating superior quality and generalization.
Jingze Shi, Yifan Wu, Bingheng Wu +4 authors
A dynamic mask sparse attention mechanism, DMA, improves long-context modeling in large language models by reducing computational complexity while maintaining information fidelity.
Fengping Tian, Chenyang Lyu, Xuanfan Ni +8 authors
A multifunctional speech synthesis system integrates voice cloning and emotion control using speaker-emotion disentanglement and rotational emotional embeddings, achieving high expressive and natural speech.
Ting Huang, Zeyu Zhang, Hao Tang
3D-R1 enhances 3D scene understanding through a high-quality synthetic dataset, reinforcement learning with GRPO, and dynamic view selection, achieving significant improvements in reasoning and generalization.
Joonmyung Choi, Sanghyeok Lee, Byungoh Ko +3 authors
Representation Shift is a training-free, model-agnostic metric that integrates token compression with FlashAttention, enabling significant speedups in video-text retrieval and video QA.
Yong Lin, Shange Tang, Bohan Lyu +17 authors
Goedel-Prover-V2, a series of open-source language models, achieves state-of-the-art performance in automated theorem proving through scaffolded data synthesis, verifier-guided self-correction, and model averaging.
Shuai Yang, Hao Li, Yilun Chen +7 authors
InstructVLA is an end-to-end vision-language-action model that enhances manipulation performance while preserving vision-language reasoning through multimodal training and mixture-of-experts adaptation.
Tongchun Zuo, Zaiyu Huang, Shuliang Ning +7 authors
DreamVVT, a two-stage framework using Diffusion Transformers and LoRA adapters, enhances video virtual try-on by leveraging unpaired human-centric data and pretrained models to preserve garment details and temporal consistency.
Mo Li, L. H. Xu, Qitai Tan +2 authors
Sculptor, a framework for Active Context Management, enhances LLM performance on long contexts by enabling proactive attention and memory control, reducing proactive interference and improving reasoning reliability.
Quan-Sheng Zeng, Yunheng Li, Qilong Wang +4 authors
A dynamic pruning framework, GlimpsePrune, improves efficiency in Large Vision-Language Models by adaptively removing irrelevant visual tokens without degrading performance.
Silin Chen, Shaoxin Lin, Xiaodong Gu +7 authors
SWE-Exp enhances software issue resolution by systematically accumulating and leveraging repair expertise from past agent experiences, improving resolution rates.
Linxin Song, Yutong Dai, Viraj Prabhu +9 authors
A multi-agent system that combines GUI control with programmatic execution improves efficiency and success in complex computer automation tasks.
Yike Zhang, Zhiyuan He, Huiqiang Jiang +4 authors
LeanK, a learning-based method, prunes unimportant key cache channels in large language models to reduce memory usage and accelerate decoding without sacrificing accuracy.
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