收录 Paper Collector 热门论文，并整理公开论点、可复现性信息和证据置信度。
Vision-language models (VLMs) commonly formulate visual grounding and detection as a coordinate-token generation problem
Interactive world models are advancing rapidly, yet existing benchmarks cover only part of the required competencies, le
Reinforcement Learning has become a standard paradigm for aligning Large Language Models with human intent and task requ
Agent skills today are hand-crafted, generated one-shot, or evolved through loosely controlled self-revision, none of wh
Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no
Multimodal Large Language Models (MLLMs) are increasingly deployed in human-facing roles where personality perception is
We introduce Lens, a 3.8B-parameter T2I model that achieves performance competitive with, and in several cases surpassin
Reinforcement learning from verifiable rewards (RLVR) has emerged as a central technique for improving the reasoning cap
Diffusion Transformers (DiTs) have become a de facto backbone of modern visual generation, and nearly every major axis o
Multimodal large language models (MLLMs) have shown remarkable capability in bridging visual perception and textual reas
Automating scientific discovery requires more than generating papers from ideas. Real research is iterative: hypotheses
Despite rapid advances in automatic speech recognition (ASR) and large audio-language models, robust recognition in real
We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of l
Without incurring significant computational overhead, train-free long video generation aims to enable foundation video g
Long-context inference in large language models is bottlenecked by the quadratic cost of full attention.
Vision-language-action models have advanced rapidly, but robot trajectories alone provide limited coverage for learning
We introduce SANA-WM, an efficient 2.6B-parameter open-source world model natively trained for one-minute generation, sy
Reinforcement learning (RL) has emerged as a central paradigm for post-training LLM agents, yet its trajectory-level rew
Real-time interactive video generation requires low-latency, streaming, and controllable rollout. Existing autoregressiv
Memory is essential for large vision-language models (LVLMs) to handle long, multimodal interactions, with two method di
The rise of personal assistant agents, e.g., OpenClaw, highlights the growing potential of large language models to supp
We present Darwin Family, a framework for training-free evolutionary merging of large language models via gradient-free
Pairwise Ranking Prompting (PRP) elicits pairwise preference judgments from an LLM, which are then aggregated into a ran
Despite rapid progress in video-capable MLLMs, we find that their apparent audio understanding in videos is often vision
Long-context modeling is becoming a core capability of modern large vision-language models (LVLMs), enabling sustained c
We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and onli
Few-step video generation has been significantly advanced by consistency distillation. However, the performance of consi
We present Qwen-Image-VAE-2.0, a suite of high-compression Variational Autoencoders (VAEs) that achieve significant adva
Reusable skills have become a core substrate for improving agent capabilities, yet most existing skill packages encode r
Recent progress in reasoning models has substantially advanced long-horizon mathematical and scientific problem solving,
Multimodal Large Language Models (MLLMs) have significantly advanced document understanding, yet current Doc-VQA evaluat
Recent large vision-language models (VLMs) remain fundamentally constrained by a persistent dichotomy: understanding and
Large language models increasingly need to accumulate and reuse historical information in long-term assistants and agent
World models enable agents to anticipate the effects of their actions by internalizing environment dynamics.
Vision-Language-Action (VLA) models have achieved strong semantic generalization for embodied policy learning, yet they
On-policy self-distillation, where a student is pulled toward a copy of itself conditioned on privileged context (e.g.,
Training deep research agents, namely systems that plan, search, evaluate evidence, and synthesize long-form reports, pu
We present Qwen-Image-2.0, an omni-capable image generation foundation model that unifies high-fidelity generation and p
Tabular Foundation Models have recently established the state of the art in supervised tabular learning, by leveraging p
As LLM-powered agents are increasingly deployed in edge-cloud environments, personalized memory has become a key enabler
Following the recent achievement of gold-medal performance on the IMO by frontier LLMs, the community is searching for t
Recent "Thinking with Video" approaches use Video Generation Models (VGMs) for visual reasoning by producing t
Test-time scaling (TTS) has become an effective approach for improving large language model performance by allocating ad
Existing Flow Matching (FM) text-to-image models suffer from two critical bottlenecks under multi-task alignment: the re
Existing multimodal search agents process target entities sequentially, issuing one tool call per entity and accumulatin
Progress in embodied intelligence increasingly depends on scalable data infrastructure. While vision and language have s
A growing body of work in cognitive science suggests that reportable conscious access is associated with \emph{global ig
Scaling Diffusion Transformers (DiTs) to hundreds of layers introduces a structural vulnerability: networks can enter a
Reinforcement learning with verifiable rewards (RLVR) has become a standard approach for large language models (LLMs) po
A persistent skill library allows language model agents to reuse successful strategies across tasks.
Deep search has become a crucial capability for frontier multimodal agents, enabling models to solve complex questions t
While Test-Time Scaling (TTS) offers a promising direction to enhance video generation without the surging costs of trai
Deep search capabilities have become an indispensable competency for frontier Large Language Model (LLM) agents, yet the
Distillation-based acceleration has become foundational for making autoregressive streaming video diffusion models pract
While Vision-Language-Action models (VLAs) have shown remarkable progress toward human-like generalist robotic policies
This report describes ARIS (Auto-Research-in-sleep), an open-source research harness for autonomous research, including
Benchmarks within the OpenClaw ecosystem have thus far evaluated exclusively assistant-level tasks, leaving the academic
Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compres
Recent progress has shown that video diffusion models (VDMs) can be repurposed for diverse multimodal graphics tasks.
Recursive or looped language models have recently emerged as a new scaling axis by iteratively refining the same model c
Reliably transferring specialized human knowledge from text into large language models remains a fundamental challenge i
Recent video foundation models demonstrate impressive visual synthesis but frequently suffer from geometric inconsistenc
Current evaluations of spatial intelligence can be systematically invalid under modern vision-language model (VLM) setti
Transformations produced by image and video generation models often evolve in a highly non-linear manner: long stretches
Individual agent capabilities have advanced rapidly through modular skills and tool integrations, yet multi-agent system
Reinforcement learning with verifiable rewards (RLVR) has become a core post-training recipe. Introducing suitable off-p
Recent advances in image generation and editing have opened new opportunities for virtual try-on. However, existing meth
Synthesizing human--object interaction (HOI) videos has broad practical value in e-commerce, digital advertising, and vi
Chain-of-Thought (CoT) reasoning has become a powerful driver of trajectory prediction in VLA-based autonomous driving,
Game development sits at the intersection of creative design and intricate software engineering, demanding the joint orc
Large language models are increasingly expected to serve as general-purpose agents that interact with external, stateful
In this work, we present Qwen3.5-Omni, the latest advancement in the Qwen-Omni model family. Representing a significant
We introduce HY-World 2.0, a multi-modal world model framework that advances our prior project HY-World 1.0.
Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 20
Spatial reasoning over three-dimensional scenes is a core capability for embodied intelligence, yet continuous model imp
Reinforcement Learning with Verifiable Rewards (RLVR) has catalyzed significant advances in the reasoning capabilities o
Language-model agent systems commonly rely on reactive prompting, in which a single instruction guides the model through
On-policy distillation (OPD) has become a core technique in the post-training of large language models, yet its training
RLVR improves reasoning in large language models, but its effectiveness is often limited by severe reward sparsity on ha
In this work, we study Human-Object Interaction Video Generation (HOIVG), which aims to synthesize high-quality human-ob
Most reward models for visual generation reduce rich human judgments to a single unexplained score, discarding the reaso
Despite the success of reinforcement learning for large language models, a common failure mode is reduced sampling diver
AI agents are expected to perform professional work across hundreds of occupational domains (from emergency department t
Recent advancements in autoregressive transformers have demonstrated remarkable potential for generating artist-quality
This technical report introduces EXAONE 4.5, the first open-weight vision language model released by LG AI Research.
Understanding objects in 3D from a single image is a cornerstone of spatial intelligence. A key step toward this goal is
Text-to-video diffusion models have enabled open-ended video synthesis, but often struggle with generating the correct n
AI agents may be able to automate your inbox, but can they automate other routine aspects of your life?
Large language model (LLM) agents such as OpenClaw rely on reusable skills to perform complex tasks, yet these skills re
In this paper, we introduce MegaStyle, a novel and scalable data curation pipeline that constructs an intra-style consis
We present DMax, a new paradigm for efficient diffusion language models (dLLMs). It mitigates error accumulation in para
Large language model (LLM) agents are increasingly built less by changing model weights than by reorganizing the runtime
Performance, the externalization of intent, emotion, and personality through visual, vocal, and temporal behavior, is wh
We introduce HY-Embodied-0.5, a family of foundation models specifically designed for real-world embodied agents.
Towards an embodied generalist for real-world interaction, Multimodal Large Language Model (MLLM) agents still suffer fr
The manufacturing sector is increasingly adopting Multimodal Large Language Models (MLLMs) to transition from simple per
A prevailing narrative in LLM post-training holds that supervised finetuning (SFT) memorizes while reinforcement learnin
RL training of multi-turn LLM agents is inherently unstable, and reasoning quality directly determines task performance.
Large language models are increasingly deployed as autonomous agents executing multi-step workflows in real-world softwa
With the rapid advancement of video understanding, existing benchmarks are becoming increasingly saturated, exposing a c
Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks. Leading KV cache compress
Image spatial editing performs geometry-driven transformations, allowing precise control over object layout and camera v
Current document parsing methods advance primarily through model architecture innovation, while systematic engineering o
Humans paint images incrementally: they plan a global layout, sketch a coarse draft, inspect, and refine details, and mo
World models have garnered significant attention as a promising research direction in artificial intelligence, yet a cle
Deep research agents (DRAs) integrate LLM reasoning with external tools. Memory systems enable DRAs to leverage historic
Video Large Language Models (VideoLLMs) have achieved strong performance on many video understanding tasks, but most exi
Selecting LLM-generated code candidates using LLM-generated tests is challenging because the tests themselves may be inc
Industrial software development across chip design, GPU optimization, and embedded systems lacks expert reasoning traces
On-policy distillation (OPD) has become a popular training paradigm in the LLM community. This paradigm selects a larger
Competitive programming remains one of the last few human strongholds in coding against AI. The best AI system to date s
Scaling generative inverse and forward rendering to real-world scenarios is bottlenecked by the limited realism and temp
Pretrained Vision Transformers (ViTs) such as DINOv2 and MAE provide generic image features that can be applied to a var
Recent streaming video understanding methods increasingly rely on complex memory mechanisms to handle long video streams
Existing video object removal methods excel at inpainting content "behind" the object and correcting appearanc
Agent skills, structured packages of procedural knowledge and executable resources that agents dynamically load at infer
While textual frequency has been validated as relevant to human cognition in reading speed, its relatedness to Large Lan
Latent space is rapidly emerging as a native substrate for language-based models. While modern systems are still commonl
Large language model (LLM)-based evolution is a promising approach for open-ended discovery, where progress requires sus
Recent studies have demonstrated that Reinforcement Learning (RL), notably Group Relative Policy Optimization (GRPO), ca
Fine-grained facial expression editing has long been limited by intrinsic semantic overlap. To address this, we construc
We introduce Intern-S1-Pro, the first one-trillion-parameter scientific multimodal foundation model.
Optical character recognition (OCR) has evolved from line-level transcription to structured document parsing, requiring
Video2GUI 是 2026-05 发布的 GUI Agent 预训练数据合成方法，用粗到精过滤策略从 5 亿条互联网视频元数据中自动提取 1200 万条 GUI 交互轨迹，构建覆盖 1500+ 应用的 WildGUI 数据集。在 Qw
MolmoAct2 是 Allen Institute for AI 2026-05 发布的视觉-语言-动作（VLA）模型，对标 Pi-05、GPT-5、Gemini Robotics ER-1.5，在 7 个仿真和现实基准上做了开放 VL
LLaDA2.0-Uni 是 Inclusion AI（蚂蚁集团 AI Lab）2026-04 发布的离散扩散大语言模型（dLLM），在原生统一框架内同时支持多模态理解和生成。架构包括完全语义离散 tokenizer + MoE-based
InCoder-32B 是 2026-03 发布的 32B 参数工业代码基础模型，统一芯片设计、GPU 内核优化、嵌入式系统、编译器优化、3D 建模五个领域的代码智能，上下文从 8K 扩展到 128K。填补通用 Code 大模型在硬件语义、
Gamma-World 是 NVIDIA 空间智能实验室 2026-05 发布的多智能体生成式世界模型，用单纯形旋转编码和稀疏枢纽注意力把跨 Agent 注意力从 O(n²) 降到 O(n)，实现 24 FPS 实时推理。arXiv:260
CARLA-Air 是 2026-03 发布的开源仿真平台，在单个虚幻引擎进程里统一高保真城市驾驶和多旋翼无人机飞行，支持 18 种传感器模态同步捕获，继承已存档的 AirSim 航空能力。适用于空地协同、具身导航、VLA、低空经济场景。a
AgentDoG 1.5 是 2026-05 发布的轻量级 Agent 安全对齐框架，专门面向 OpenClaw、Codex 等开放世界 Agent 的跨环境执行风险。用约 1000 个样本训练 0.8B/2B/4B/8B 多个参数规模变体