Trending Papers

The Qwen3-TTS series presents advanced multilingual text-to-speech models with voice cloning and controllable speech generation capabilities, utilizing dual-track LM architecture and specialized speech tokenizers for efficient streaming synthesis.

A Vision-Language-Action model trained on extensive real-world robotic data demonstrates superior performance and generalization across multiple platforms while offering enhanced efficiency through optimized training infrastructure.

LingBot-Depth is a depth completion model that uses visual context to refine depth maps through masked depth modeling and automated data curation for improved spatial perception in robotics and autonomous systems.

UltraRAG is a comprehensive RAG toolkit that automates knowledge adaptation across the entire workflow while providing a user-friendly interface for non-coding deployment.

VibeVoice synthesizes long-form multi-speaker speech using next-token diffusion and a highly efficient continuous speech tokenizer, achieving superior performance and fidelity.

A suite of open-source music foundation models is introduced, featuring components for audio-text alignment, lyric recognition, music coding, and large language model-based song generation with controllable attributes and scalable parameterization.

To support reliable long-term interaction in complex environments, LLM agents require memory systems that efficiently manage historical experiences. Existing approaches either retain full interaction histories via passive context extension, leading to substantial redundancy, or rely on iterative reasoning to filter noise, incurring high token costs. To address this challenge, we introduce SimpleMem, an efficient memory framework based on semantic lossless compression. We propose a three-stage pipeline designed to maximize information density and token utilization: (1) Semantic Structured Compression, which applies entropy-aware filtering to distill unstructured interactions into compact, multi-view indexed memory units; (2) Recursive Memory Consolidation, an asynchronous process that integrates related units into higher-level abstract representations to reduce redundancy; and (3) Adaptive Query-Aware Retrieval, which dynamically adjusts retrieval scope based on query complexity to construct precise context efficiently. Experiments on benchmark datasets show that our method consistently outperforms baseline approaches in accuracy, retrieval efficiency, and inference cost, achieving an average F1 improvement of 26.4% while reducing inference-time token consumption by up to 30-fold, demonstrating a superior balance between performance and efficiency. Code is available at https://github.com/aiming-lab/SimpleMem.

PagedAttention algorithm and vLLM system enhance the throughput of large language models by efficiently managing memory and reducing waste in the key-value cache.

MinerU2.5, a 1.2B-parameter document parsing vision-language model, achieves state-of-the-art recognition accuracy with computational efficiency through a coarse-to-fine parsing strategy.

MinerU is an open-source tool that enhances document content extraction using fine-tuned models and pre/postprocessing rules across diverse document types.

Agentic reasoning redefines large language models as autonomous agents capable of planning, acting, and learning through continuous interaction in dynamic environments across single-agent and multi-agent frameworks.

SmolDocling is a compact vision-language model that performs end-to-end document conversion with robust performance across various document types using 256M parameters and a new markup format.

AdaReasoner enables multimodal models to learn tool usage as a general reasoning skill through scalable data curation, reinforcement learning for tool selection, and adaptive learning mechanisms that improve performance on complex visual reasoning tasks.

Audio Language Models (ALM) have emerged as the dominant paradigm for speech and music generation by representing audio as sequences of discrete tokens. Yet, unlike text tokens, which are invertible, audio tokens are extracted from lossy codecs with a limited bitrate. As a consequence, increasing audio quality requires generating more tokens, which imposes a trade-off between fidelity and computational cost. We address this issue by studying Continuous Audio Language Models (CALM). These models instantiate a large Transformer backbone that produces a contextual embedding at every timestep. This sequential information then conditions an MLP that generates the next continuous frame of an audio VAE through consistency modeling. By avoiding lossy compression, CALM achieves higher quality at lower computational cost than their discrete counterpart. Experiments on speech and music demonstrate improved efficiency and fidelity over state-of-the-art discrete audio language models, facilitating lightweight, high-quality audio generation. Samples are available at https://continuous-audio-language-models.github.io

AI agents face safety and security challenges from autonomous tool use and environmental interactions, requiring advanced guardrail frameworks for risk diagnosis and transparent monitoring.

Agent Lightning is a flexible RL framework for training LLMs in various agents, using a hierarchical RL algorithm and decoupling execution from training to handle complex interactions.

Z-Image, a 6B-parameter Scalable Single-Stream Diffusion Transformer (S3-DiT) model, achieves high-performance image generation with reduced computational cost, offering sub-second inference and compatibility with consumer hardware.

Mem0, a memory-centric architecture with graph-based memory, enhances long-term conversational coherence in LLMs by efficiently extracting, consolidating, and retrieving information, outperforming existing memory systems in terms of accuracy and computational efficiency.

The study reveals that in text-to-image generation, CFG Augmentation is the primary driver of few-step distillation in Distribution Matching Distillation (DMD), while the distribution matching term acts as a regularizer.

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.

Finite Scalar Quantization with improved activation mapping enables unified modeling of discrete and continuous image generation approaches, revealing optimal representation balance and performance characteristics.

Agentic coding tools receive goals written in natural language as input, break them down into specific tasks, and write or execute the actual code with minimal human intervention. Central to this process are agent context files ("READMEs for agents") that provide persistent, project-level instructions. In this paper, we conduct the first large-scale empirical study of 2,303 agent context files from 1,925 repositories to characterize their structure, maintenance, and content. We find that these files are not static documentation but complex, difficult-to-read artifacts that evolve like configuration code, maintained through frequent, small additions. Our content analysis of 16 instruction types shows that developers prioritize functional context, such as build and run commands (62.3%), implementation details (69.9%), and architecture (67.7%). We also identify a significant gap: non-functional requirements like security (14.5%) and performance (14.5%) are rarely specified. These findings indicate that while developers use context files to make agents functional, they provide few guardrails to ensure that agent-written code is secure or performant, highlighting the need for improved tooling and practices.

HunyuanVideo 1.5 is a lightweight video generation model with state-of-the-art visual quality and motion coherence, using a DiT architecture with SSTA and an efficient video super-resolution network.

Self-refining video sampling improves motion coherence and physics alignment by using a pre-trained video generator as its own denoising autoencoder for iterative refinement with uncertainty-aware region selection.

Agents that evolve tools through continuous interaction and feedback can adapt to dynamic environments and transfer knowledge across domains more effectively than traditional systems.

FastNeRF enables high-speed rendering of photorealistic 3D environments by factorizing radiance maps for efficient pixel value estimation.

A new solver, DPM-Solver, accelerates sampling from diffusion probabilistic models by analytically solving the diffusion ordinary differential equations, achieving high-quality results with fewer function evaluations compared to existing methods.

Rectified flow is a simple ODE-based method for efficient distribution transport and tasks like generative modeling and domain transfer, achieving high-quality results with minimal computational cost.

MemOS, a memory operating system for Large Language Models, addresses memory management challenges by unifying plaintext, activation-based, and parameter-level memories, enabling efficient storage, retrieval, and continual learning.

Chroma 1.0 enables real-time spoken dialogue with personalized voice cloning through discrete speech representations and interleaved text-audio token scheduling.

A multi-agent framework using large language models for stock trading simulates real-world trading firms, improving performance metrics like cumulative returns and Sharpe ratio.

LlamaFactory is a unified framework enabling efficient fine-tuning of large language models across various tasks using a web-based user interface.

A self-supervised framework optimizes prompts for both closed and open-ended tasks by evaluating LLM outputs without external references, reducing costs and required data.

Test-time training enables AI systems to discover optimal solutions for specific scientific problems through continual learning focused on individual challenges rather than generalization.

We present MiroThinker v1.0, an open-source research agent designed to advance tool-augmented reasoning and information-seeking capabilities. Unlike previous agents that only scale up model size or context length, MiroThinker explores interaction scaling at the model level, systematically training the model to handle deeper and more frequent agent-environment interactions as a third dimension of performance improvement. Unlike LLM test-time scaling, which operates in isolation and risks degradation with longer reasoning chains, interactive scaling leverages environment feedback and external information acquisition to correct errors and refine trajectories. Through reinforcement learning, the model achieves efficient interaction scaling: with a 256K context window, it can perform up to 600 tool calls per task, enabling sustained multi-turn reasoning and complex real-world research workflows. Across four representative benchmarks-GAIA, HLE, BrowseComp, and BrowseComp-ZH-the 72B variant achieves up to 81.9%, 37.7%, 47.1%, and 55.6% accuracy respectively, surpassing previous open-source agents and approaching commercial counterparts such as GPT-5-high. Our analysis reveals that MiroThinker benefits from interactive scaling consistently: research performance improves predictably as the model engages in deeper and more frequent agent-environment interactions, demonstrating that interaction depth exhibits scaling behaviors analogous to model size and context length. These findings establish interaction scaling as a third critical dimension for building next-generation open research agents, complementing model capacity and context windows.

OpenDevin is a platform for developing AI agents that interact with the world by writing code, using command lines, and browsing the web, with support for multiple agents and evaluation benchmarks.

ARK is an open-source Python-first framework that integrates modern imitation-learning algorithms and seamless simulation-physical robot interactions to simplify robotics development and deployment.

World Craft enables non-expert users to create executable and visualizable AI environments through textual descriptions by combining structured scaffolding and multi-agent intent analysis.

SWE-Pruner is a self-adaptive context pruning framework for coding agents that uses task-aware pruning to reduce token usage while maintaining performance.

Conditional memory via Engram module enhances Transformer models by enabling efficient knowledge lookup and improving reasoning capabilities through optimized sparsity allocation.

RebuttalAgent is a multi-agent framework that reframes rebuttal generation as an evidence-centric planning task, improving coverage, faithfulness, and strategic coherence in academic peer review.

LTX-2 is an open-source audiovisual diffusion model that generates synchronized video and audio content using a dual-stream transformer architecture with cross-modal attention and classifier-free guidance.

SHARP synthesizes photorealistic views from a single image using a 3D Gaussian representation, achieving state-of-the-art results with rapid processing.

Enhancements to the AgentScope platform improve scalability, efficiency, and ease of use for large-scale multi-agent simulations through distributed mechanisms, flexible environments, and user-friendly tools.

AgentScope enhances agentic applications by providing flexible tool-based interactions, unified interfaces, and advanced infrastructure based on the ReAct paradigm, supporting efficient and safe development and deployment.

The PyTorch distributed data parallel module optimizes large-scale model training using techniques like gradient bucketing, computation-communication overlap, and selective synchronization to achieve near-linear scalability.

PyTorch Fully Sharded Data Parallel (FSDP) enables efficient and scalable training of large models across hardware configurations.

Prometheus, a multi-agent system using a knowledge graph and DeepSeek-V3, resolves real-world issues across multiple programming languages with high efficiency.

Flawed-Aware Policy Optimization (FAPO) enhances reinforcement learning with verifiable rewards by penalizing flawed-positive rollouts, improving reasoning capability and training stability in large language models.

GigaBrain-0, a VLA foundation model, uses world model-generated data to enhance cross-task generalization and policy robustness, improving real-world performance on complex manipulation tasks.