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。
Wei Huang, Yi Ge, Shuai Yang +11 authors
QeRL, a quantization-enhanced reinforcement learning framework, accelerates RL training for large language models by combining NVFP4 quantization with Low-Rank Adaptation and an Adaptive Quantization Noise mechanism, achieving significant speedups and improved performance.
50 篇论文 · 按点赞排序
Boyang Zheng, Nanye Ma, Shengbang Tong +1 authors
Replacing VAEs with pretrained representation encoders in Diffusion Transformers enhances generative quality and convergence speed without auxiliary losses.
Fuhao Li, Wenxuan Song, Han Zhao +5 authors
Vision-language-action (VLA) models have recently shown strong potential in enabling robots to follow language instructions and execute precise actions. However, most VLAs are built upon vision-language models pretrained solely on 2D data, which lack accurate spatial awareness and hinder their ability to operate in the 3D physical world. Existing solutions attempt to incorporate explicit 3D sensor inputs such as depth maps or point clouds, but these approaches face challenges due to sensor noise, hardware heterogeneity, and incomplete depth coverage in existing datasets. Alternative methods that estimate 3D cues from 2D images also suffer from the limited performance of depth estimators.We propose Spatial Forcing (SF), a simple yet effective alignment strategy that implicitly forces VLA models to develop spatial comprehension capabilities without relying on explicit 3D inputs or depth estimators. SF aligns intermediate visual embeddings of VLAs with geometric representations produced by pretrained 3D foundation models. By enforcing alignment at intermediate layers, SF guides VLAs to encode richer spatial representations that enhance action precision.Extensive experiments in simulation and real-world environments demonstrate that SF achieves state-of-the-art results, surpassing both 2D- and 3D-based VLAs. SF further accelerates training by up to 3.8x and improves data efficiency across diverse robotic tasks. Project page is at https://spatial-forcing.github.io/
Suwhan Choi, Jaeyoon Jung, Haebin Seong +7 authors
D2E framework uses desktop interactions to pretrain embodied AI, achieving high success rates in physical manipulation and navigation tasks.
Francesco Capuano, Caroline Pascal, Adil Zouitine +2 authors
Robot learning transitions from model-based to data-driven methods, leveraging reinforcement learning and behavioral cloning to develop versatile, language-conditioned models for diverse tasks and robot types.
Cheng Cui, Ting Sun, Suyin Liang +15 authors
PaddleOCR-VL, a vision-language model combining NaViT-style dynamic resolution and ERNIE, achieves state-of-the-art performance in document parsing and element recognition with high efficiency.
Kang Liao, Size Wu, Zhonghua Wu +6 authors
Puffin, a unified multimodal model, integrates language regression and diffusion-based generation to enhance camera-centric spatial understanding and generation by treating camera parameters as language.
Elisei Rykov, Kseniia Petrushina, Maksim Savkin +6 authors
PsiloQA, a multilingual dataset with span-level hallucinations, enhances hallucination detection in large language models across 14 languages using an automated pipeline and encoder-based models.
Jiachen Lei, Keli Liu, Julius Berner +4 authors
Pixel-space generative models are often more difficult to train and generally underperform compared to their latent-space counterparts, leaving a persistent performance and efficiency gap. In this paper, we introduce a novel two-stage training framework that closes this gap for pixel-space diffusion and consistency models. In the first stage, we pre-train encoders to capture meaningful semantics from clean images while aligning them with points along the same deterministic sampling trajectory, which evolves points from the prior to the data distribution. In the second stage, we integrate the encoder with a randomly initialized decoder and fine-tune the complete model end-to-end for both diffusion and consistency models. Our training framework demonstrates strong empirical performance on ImageNet dataset. Specifically, our diffusion model reaches an FID of 2.04 on ImageNet-256 and 2.35 on ImageNet-512 with 75 number of function evaluations (NFE), surpassing prior pixel-space methods by a large margin in both generation quality and efficiency while rivaling leading VAE-based models at comparable training cost. Furthermore, on ImageNet-256, our consistency model achieves an impressive FID of 8.82 in a single sampling step, significantly surpassing its latent-space counterpart. To the best of our knowledge, this marks the first successful training of a consistency model directly on high-resolution images without relying on pre-trained VAEs or diffusion models.
Guanting Dong, Licheng Bao, Zhongyuan Wang +11 authors
AEPO, an agentic RL algorithm, addresses entropy-related challenges in web agent training, enhancing performance and stability across various datasets.
Chenghao Xiao, Hou Pong Chan, Hao Zhang +3 authors
Recent multimodal embedding approaches leveraging multimodal large language models (MLLMs) fine-tuned with contrastive learning (CL) have shown promising results, yet the underlying reasons behind their superiority remain underexplored. This work argues that a crucial advantage of MLLM-based approaches stems from implicit cross-modal alignment achieved during generative pretraining, where the language decoder learns to exploit multimodal signals within a shared representation space for generating unimodal outputs. Through analysis of anisotropy and kernel similarity structure, we empirically confirm that latent alignment emerges within MLLM representations, allowing CL to serve as a lightweight refinement stage. Leveraging this insight, we propose a Language-Centric Omnimodal Embedding framework, termed LCO-Emb. Extensive experiments across diverse backbones and benchmarks demonstrate its effectiveness, achieving state-of-the-art performance across modalities. Furthermore, we identify a Generation-Representation Scaling Law (GRSL), showing that the representational capabilities gained through contrastive refinement scales positively with the MLLM's generative capabilities. This suggests that improving generative abilities evolves as an effective paradigm for enhancing representation quality. We provide a theoretical explanation of GRSL, which formally links the MLLM's generative quality to the upper bound on its representation performance, and validate it on a challenging, low-resource visual-document retrieval task, showing that continual generative pretraining before CL can further enhance the potential of a model's embedding capabilities. Codes, models, and resources are available at https://github.com/LCO-Embedding/LCO-Embedding.
Enze Zhang, Jiaying Wang, Mengxi Xiao +7 authors
A new evaluation framework, DITING, and a reasoning-driven multi-agent evaluation framework, AgentEval, are introduced to assess the quality of web novel translations, revealing that Chinese-trained LLMs outperform larger foreign models.
Minjun Kim, Hyeonseok Lim, Hangyeol Yoo +10 authors
A large-scale investigation into constructing a fully open bilingual LLM for Korean using synthetic data demonstrates that such data can sustain pretraining and achieve performance comparable to multilingual baselines.
Hengyuan Xu, Wei Cheng, Peng Xing +8 authors
A diffusion-based model addresses copy-paste artifacts in text-to-image generation by using a large-scale paired dataset and a contrastive identity loss to balance identity fidelity and variation.
Zirui Guo, Xubin Ren, Lingrui Xu +2 authors
RAG-Anything is a unified framework that enhances multimodal knowledge retrieval by integrating cross-modal relationships and semantic matching, outperforming existing methods on complex benchmarks.
Zichen Wen, Yiyu Wang, Chenfei Liao +10 authors
Alpha-Service, a unified framework for proactive AI assistance, uses a multi-agent system on AI glasses to detect service opportunities and provide timely, personalized assistance.
Xinyang Li, Tengfei Wang, Zixiao Gu +3 authors
FlashWorld generates 3D scenes from single images or text prompts quickly and with high quality by combining MV-oriented and 3D-oriented generation methods.
Haiwen Diao, Mingxuan Li, Silei Wu +6 authors
NEO, a novel family of native Vision-Language Models, addresses fundamental constraints and integrates vision and language within a unified framework, achieving competitive performance with limited data.
Zhenyu Liu, Yunxin Li, Xuanyu Zhang +13 authors
UniMoE-Audio, a unified speech and music generation model using a Dynamic-Capacity Mixture-of-Experts framework, addresses data imbalance and task conflicts, achieving state-of-the-art performance and enhanced cross-domain synergy.
Xun Wu, Shaohan Huang, Wenhui Wang +4 authors
BitNet Distillation fine-tunes large language models to 1.58-bit precision using SubLN, multi-head attention distillation, and continual pre-training, achieving comparable performance with significant memory and inference speed improvements.
Yi Zhang, Bolin Ni, Xin-Sheng Chen +7 authors
A new dataset and pipeline for data curation improve the performance of fully open multimodal large language models, achieving state-of-the-art results competitive with semi-open models.
Yang Li, Zhichen Dong, Yuhan Sun +9 authors
Attention mechanisms in LLMs are analyzed to reveal reasoning patterns, leading to novel RL strategies that improve performance by focusing on critical tokens.
Meiqi Wu, Jiashu Zhu, Xiaokun Feng +7 authors
ImagerySearch, a prompt-guided adaptive test-time search strategy, enhances video generation in imaginative scenarios by dynamically adjusting search spaces and reward functions, outperforming existing methods on a new benchmark, LDT-Bench.
Qing Jiang, Junan Huo, Xingyu Chen +6 authors
Object detection has long been dominated by traditional coordinate regression-based models, such as YOLO, DETR, and Grounding DINO. Although recent efforts have attempted to leverage MLLMs to tackle this task, they face challenges like low recall rate, duplicate predictions, coordinate misalignment, etc. In this work, we bridge this gap and propose Rex-Omni, a 3B-scale MLLM that achieves state-of-the-art object perception performance. On benchmarks like COCO and LVIS, Rex-Omni attains performance comparable to or exceeding regression-based models (e.g., DINO, Grounding DINO) in a zero-shot setting. This is enabled by three key designs: 1) Task Formulation: we use special tokens to represent quantized coordinates from 0 to 999, reducing the model's learning difficulty and improving token efficiency for coordinate prediction; 2) Data Engines: we construct multiple data engines to generate high-quality grounding, referring, and pointing data, providing semantically rich supervision for training; \3) Training Pipelines: we employ a two-stage training process, combining supervised fine-tuning on 22 million data with GRPO-based reinforcement post-training. This RL post-training leverages geometry-aware rewards to effectively bridge the discrete-to-continuous coordinate prediction gap, improve box accuracy, and mitigate undesirable behaviors like duplicate predictions that stem from the teacher-guided nature of the initial SFT stage. Beyond conventional detection, Rex-Omni's inherent language understanding enables versatile capabilities such as object referring, pointing, visual prompting, GUI grounding, spatial referring, OCR and key-pointing, all systematically evaluated on dedicated benchmarks. We believe that Rex-Omni paves the way for more versatile and language-aware visual perception systems.
Qinglin Zhu, Yizhen Yao, Runcong Zhao +7 authors
Latent Refinement Decoding (LRD) improves parallel sequence generation by maintaining global consistency and iterative refinement, enhancing accuracy and reducing latency.
Ruyi Xu, Guangxuan Xiao, Yukang Chen +4 authors
StreamingVLM is a real-time vision-language model that efficiently processes infinite video streams using a compact KV cache and supervised fine-tuning, achieving high performance on long videos and diverse benchmarks.
Qiguang Chen, Zheng Yan, Mingda Yang +10 authors
AutoPR, a multi-agent framework, automates the promotion of research papers by transforming them into engaging public content, significantly improving engagement metrics compared to direct LLM pipelines.
Yuyao Ge, Lingrui Mei, Zenghao Duan +12 authors
The advancement of large language models (LLMs) has catalyzed a paradigm shift from code generation assistance to autonomous coding agents, enabling a novel development methodology termed "Vibe Coding" where developers validate AI-generated implementations through outcome observation rather than line-by-line code comprehension. Despite its transformative potential, the effectiveness of this emergent paradigm remains under-explored, with empirical evidence revealing unexpected productivity losses and fundamental challenges in human-AI collaboration. To address this gap, this survey provides the first comprehensive and systematic review of Vibe Coding with large language models, establishing both theoretical foundations and practical frameworks for this transformative development approach. Drawing from systematic analysis of over 1000 research papers, we survey the entire vibe coding ecosystem, examining critical infrastructure components including LLMs for coding, LLM-based coding agent, development environment of coding agent, and feedback mechanisms. We first introduce Vibe Coding as a formal discipline by formalizing it through a Constrained Markov Decision Process that captures the dynamic triadic relationship among human developers, software projects, and coding agents. Building upon this theoretical foundation, we then synthesize existing practices into five distinct development models: Unconstrained Automation, Iterative Conversational Collaboration, Planning-Driven, Test-Driven, and Context-Enhanced Models, thus providing the first comprehensive taxonomy in this domain. Critically, our analysis reveals that successful Vibe Coding depends not merely on agent capabilities but on systematic context engineering, well-established development environments, and human-agent collaborative development models.
Yumin Choi, Dongki Kim, Jinheon Baek +1 authors
Multimodal Prompt Optimizer (MPO) extends prompt optimization to handle multiple data types, improving performance over text-only methods in various applications.
Senyu Fei, Siyin Wang, Junhao Shi +10 authors
State-of-the-art Visual-Language-Action models show high benchmark scores but are brittle to various perturbations, particularly in camera viewpoints and robot initial states, and often ignore language instructions.
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