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。
Adrian Kosowski, Przemysław Uznański, Jan Chorowski +2 authors
BDH, a biologically inspired Large Language Model, combines scale-free network architecture with Hebbian learning to achieve Transformer-like performance while maintaining interpretability.
50 篇论文 · 按点赞排序
Alexia Jolicoeur-Martineau
Tiny Recursive Model (TRM) achieves high generalization on complex puzzle tasks using a small, two-layer network with minimal parameters, outperforming larger language models.
Kai Zhang, Xiangchao Chen, Bo Liu +27 authors
Early experience, using agent-generated interaction data without reward signals, improves policy effectiveness and generalization, serving as a bridge between imitation learning and reinforcement learning.
Rui-Jie Zhu, Zixuan Wang, Kai Hua +30 authors
LoopLM, a family of pre-trained Looped Language Models, enhances reasoning by integrating iterative computation and entropy regularization during pre-training, achieving superior performance with better knowledge manipulation.
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.
Yujia Zhang, Xiaoyang Wu, Yixing Lao +4 authors
Concerto, a minimalist model combining 3D self-distillation and 2D-3D joint embedding, achieves superior spatial feature learning and outperforms existing models in scene understanding and open-world perception.
Zijian Wu, Xiangyan Liu, Xinyuan Zhang +12 authors
MCPMark is a comprehensive benchmark for evaluating MCP use in real-world workflows, featuring diverse tasks that require richer interactions with the environment, and reveals that current LLMs perform poorly on these tasks.
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.
Zhi Zhou, Yuhao Tan, Zenan Li +4 authors
A theoretical framework for sampling-based test-time scaling in large language models reveals limitations in self-consistency and perplexity, and introduces RPC to improve reasoning performance and reduce sampling costs.
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.
Fang Wu, Weihao Xuan, Heli Qi +4 authors
DeepSearch integrates Monte Carlo Tree Search into RLVR training to enhance exploration and credit assignment, achieving state-of-the-art performance with reduced computational cost.
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.
Qinsi Wang, Bo Liu, Tianyi Zhou +6 authors
Vision-Zero is a domain-agnostic framework that enhances vision-language models through self-improvement in competitive visual games, using Iterative Self-Play Policy Optimization and achieving state-of-the-art performance without human annotation.
Kimi Team, Yu Zhang, Zongyu Lin +57 authors
Kimi Linear, a hybrid linear attention architecture, outperforms full attention in various scenarios with improved efficiency and performance, using Kimi Delta Attention and Multi-Head Latent Attention.
Qizheng Zhang, Changran Hu, Shubhangi Upasani +10 authors
ACE, a framework for adaptive context engineering, enhances LLM applications by preserving detailed knowledge through structured updates, outperforming baselines in agent and domain-specific tasks with reduced adaptation costs.
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.
Zhaoyang Yu, Jiayi Zhang, Huixue Su +10 authors
ReCode, a recursive code generation paradigm, unifies high-level planning and low-level action in a single representation, enhancing decision granularity and data efficiency in LLM-based agents.
Yonghao Zhuang, Junda Chen, Bo Pang +6 authors
CAD, a technique for long-context large language model training, improves throughput and balance by decoupling and distributing core attention computations.
Zhichao Wang, Dongyang Ma, Xinting Huang +6 authors
AutoDeco, a novel architecture, enables end-to-end generation by dynamically predicting decoding parameters, outperforming default strategies and achieving performance close to an oracle-tuned baseline.
Zeyu Zhu, Kevin Qinghong Lin, Mike Zheng Shou
PaperTalker is a multi-agent framework that automates academic presentation video generation by integrating slide generation, layout refinement, subtitling, speech synthesis, and talking-head rendering, outperforming existing methods.
Yufeng Cui, Honghao Chen, Haoge Deng +20 authors
Emu3.5, a large-scale multimodal world model, predicts next states in vision and language, enhanced with reinforcement learning and Discrete Diffusion Adaptation for efficient inference, achieving strong performance in various multimodal tasks.
Jizhan Fang, Xinle Deng, Haoming Xu +9 authors
LightMem, a memory system inspired by human memory, enhances LLMs by efficiently managing historical interaction information, improving accuracy and reducing computational costs.
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.
Ling Team, Bin Han, Caizhi Tang +25 authors
The Ring-linear model series, including Ring-mini-linear-2.0 and Ring-flash-linear-2.0, uses a hybrid architecture combining linear and softmax attention to reduce inference costs and improve training efficiency.
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.
Shaolei Zhang, Ju Fan, Meihao Fan +2 authors
DeepAnalyze-8B, an agentic LLM, autonomously completes the data science pipeline from raw data to research reports using curriculum-based training and data-grounded trajectory synthesis.
Zhuofeng Li, Haoxiang Zhang, Seungju Han +6 authors
AgentFlow, a trainable agentic framework with in-the-flow optimization, enhances reasoning in large language models by coordinating specialized modules and outperforms top baselines across various tasks.
Xiangyu Zhao, Junming Lin, Tianhao Liang +11 authors
Existing Multimodal Large Language Models show performance deficits in long-chain reflective reasoning, which is addressed by developing MM-HELIX-100K and Adaptive Hybrid Policy Optimization, leading to improved accuracy and generalization.
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