xClean Tools

AI Papers — Daily Top 5

UPDATED 2026-09-05 13:05 PDT

2026-07-25 · SATURDAY · 11:23 PDT

  1. 1 NVIDIA-labs OO Agents: Native Python Object-Oriented Agents 19 UPVOTES · PAUL FURGALE ET AL. · NVIDIA · ARXIV 2607.20709 Nvidia proposes NOOA, a framework in which an agent is simply a Python object: methods are its actions, fields its state, docstrings its prompts and type annotations its contracts. A method whose body is only an ellipsis gets completed at runtime by an LLM loop, while normal method bodies stay deterministic Python. The aim is to collapse prompt templates, tool schemas and workflow graphs into one artifact.
  2. 2 Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction 19 UPVOTES · TENCENT WORKBUDDY BENCH TEAM ET AL. · TENCENT · ARXIV 2607.20911 Tencent released an evaluation suite for coding agents spanning four work domains: code, web, office and security. Rather than reusing public issue text, each task is reverse-engineered from a real commit, pull request or business scenario and rewritten as a short colloquial request, so the prompt cannot be recovered from training data. The report also gives a cross-model leaderboard.
  3. 3 Color Pass-Through via Camera-Display Coupling 18 UPVOTES · RUIKANG LI ET AL. · MMLAB-CUHK · ARXIV 2607.12746 A scene shot on a phone and shown on the same phone's screen still looks noticeably off in color, brightness and contrast. The authors trace the gap to pipelines that calibrate the camera and the display separately and then bridge them with low-dimensional color transforms, which creates information bottlenecks and compounding error. They propose treating capture and display as one coupled system instead.
  4. 4 SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals for Efficient Video Generation 18 UPVOTES · JUNSONG CHEN ET AL. · NVIDIA · ARXIV 2607.21553 SANA-Video 2.0 is a hybrid video diffusion transformer at 5B and 14B scale that generates up to 720p video on a single GPU. It mixes gated linear attention for the bulk of token mixing with periodic gated-softmax anchors at a 3:1 ratio, restoring the full-rank token interactions pure linear attention loses. The claim is full-softmax quality with linear attention's long-sequence scaling.
  5. 5 LLMs Get Lost in Evolving User Intent 17 UPVOTES · JIHOON TACK ET AL. · MICROSOFT RESEARCH · ARXIV 2607.20734 Real users rarely state what they want upfront; they disclose, revise and reshape it as a conversation unfolds. Yet models are still mostly evaluated and trained on single-turn, fully-specified prompts. Microsoft Research introduces a framework that turns fully-specified tasks into evolving ones to test how well models track intent that keeps moving.

2026-07-24 · FRIDAY · 15:44 PDT

  1. 1 AREX: Towards a Recursively Self-Improving Agent for Deep Research 115 UPVOTES · SHUQI LU ET AL. · BEIJING ACADEMY OF ARTIFICIAL INTELLIGENCE · ARXIV 2607.21461 AREX proposes an agent for deep research that recursively improves itself, refining its own research strategies over successive iterations rather than staying fixed. It drew the day's most upvotes on Hugging Face by a wide margin. The work is from the Beijing Academy of Artificial Intelligence.
  2. 2 ReferTrack: Referring Then Tracking for Embodied Visual Tracking 43 UPVOTES · HANJING YE ET AL. · TENCENT · ARXIV 2607.20061 ReferTrack tackles embodied visual tracking by first resolving a natural-language reference to the target, then tracking it, rather than tracking blindly. The two-stage design aims to make an embodied agent follow the object a user actually means. The paper is from Tencent.
  3. 3 K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs 40 UPVOTES · HAO LIANG ET AL. · PEKING UNIVERSITY · ARXIV 2605.09635 K12-KGraph introduces a knowledge graph aligned to K-12 curricula, built to benchmark and train educational large language models against structured grade-level content. It targets the gap between general LLM knowledge and what students are actually taught. The work comes from Peking University.
  4. 4 Visual Contrastive Self-Distillation 39 UPVOTES · YIJUN LIANG ET AL. · UNIVERSITY OF MARYLAND COLLEGE PARK · ARXIV 2607.21556 Visual Contrastive Self-Distillation presents a self-supervised vision method that combines contrastive learning with self-distillation, letting a model learn visual representations by teaching itself. The approach aims to improve representation quality without extra labels. The paper is from the University of Maryland, College Park.
  5. 5 Show, Don't Tell: Evaluating Spatial Cognition in Generative Pixels Rather Than LLM Text 34 UPVOTES · XU WANG ET AL. · ZJU-OMNIAI · ARXIV 2607.21072 This paper argues spatial cognition should be evaluated by what generative models draw, not what LLMs say, and builds a benchmark that scores spatial understanding directly in generated pixels. It contends text-based tests miss capabilities and failures visible only in imagery. The work is from ZJU-OmniAI.

2026-07-23 · THURSDAY · 23:34 PDT

  1. 1 Generative World Renderer at the Speed of Play 67 UPVOTES · GUIXU LIN ET AL. · ALAYA LAB · ARXIV 2607.18703 The paper introduces a generative world renderer that produces game-like visual worlds fast enough to run at interactive play speed. It frames real-time world generation as a rendering problem a single model can drive rather than a slow offline pipeline. The work drew the day's most upvotes on Hugging Face.
  2. 2 Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing 60 UPVOTES · XINJIE ZHANG ET AL. · MICROSOFT · ARXIV 2607.19064 Microsoft researchers present Mage-Flow, a compact foundation model for image generation and instruction-based editing that operates at native resolution instead of upscaling from a fixed low-resolution latent. The paper emphasizes efficiency at full resolution over raw parameter count. It was among the day's most-upvoted submissions.
  3. 3 SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD 54 UPVOTES · DONGFANG LI ET AL. · ARXIV 2607.20145 The paper documents full-parameter post-training of the DeepSeek-V4 model family on Huawei's Ascend SuperPOD, a large-scale run outside the usual Nvidia stack. It details the systems engineering needed to fine-tune frontier-scale models on Ascend hardware. It ranked among the day's top papers.
  4. 4 Subliminal Clocks: Latent Time Modelling in Diffusion Language Models 36 UPVOTES · MAXIMO EDUARDO RULLI ET AL. · SAPIENZA UNIVERSITY OF ROME · ARXIV 2607.01774 This paper studies how diffusion language models implicitly track progress through generation, proposing latent 'subliminal clocks' that encode timing inside the denoising process. It offers a lens on how such models order their output without explicit autoregression. The work came from Sapienza University of Rome.
  5. 5 Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning 31 UPVOTES · JUNYAO YANG ET AL. · TENCENT HUNYUAN · ARXIV 2607.18722 Tencent Hunyuan researchers propose a staleness-adaptive trust region to stabilize asynchronous reinforcement learning, where updates computed on outdated policy copies can destabilize training. The method scales each update's trust region by how stale it is. It aims at faster yet more stable large-scale RL.

2026-07-22 · WEDNESDAY

  1. 1 RynnBrain 1.1: Towards More Capable and Generalizable Embodied Foundation Model 187 UPVOTES · KEHAN LI ET AL. · DAMO ACADEMY · ARXIV 2607.17977 DAMO Academy presents RynnBrain 1.1, a family of embodied foundation models spanning 2B to 122B parameters, aimed at more capable and generalizable control for physical agents.
  2. 2 ABot-World-0: Infinite Interactive World Rollout on a Single Desktop GPU 170 UPVOTES · FAN JIANG ET AL. · ALIBABA AMAP CV LAB · ARXIV 2607.19191 Alibaba's AMAP CV Lab introduces ABot-World-0, an action-conditioned video world model for real-time, long-horizon interactive rollout - a step toward world models robots can plan against.
  3. 3 DataFlow-Harness: A Grounded Code-Agent Platform for Constructing Editable LLM Data Pipelines 122 UPVOTES · RUNMING HE ET AL. · PEKING UNIVERSITY · ARXIV 2607.16617 From Peking University, DataFlow-Harness is a grounded code-agent platform for automating data-processing workflows, giving agents an executable, verifiable substrate instead of free-form code.
  4. 4 SWE-Pruner Pro: The Coder LLM Already Knows What to Prune 71 UPVOTES · YUHANG WANG ET AL. · BYTEDANCE · ARXIV 2607.18213 ByteDance's SWE-Pruner Pro prunes long context for coding agents by using the coder model's own signals about what matters, cutting tokens while keeping the context an agent actually needs.
  5. 5 Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers 67 UPVOTES · MAOHUA LI ET AL. · RTP-LLM · ARXIV 2607.19139 This paper studies text-to-image diffusion transformers and argues the template tokens they process act as implicit semantic registers, offering a new handle on how DiTs bind text to image.

2026-07-21 · TUESDAY

  1. 1 TimeLens2: Generalist Video Temporal Grounding with Multimodal LLMs 137 UPVOTES · YUHAN ZHU ET AL. · MULTIMEDIA COMPUTING GROUP-NANJING UNIVERSITY · ARXIV 2607.17423 Video MLLMs can describe what happens in a video but rarely when the supporting evidence occurs. TimeLens2 targets generalist temporal grounding - locating the moments that back an answer across video tasks.
  2. 2 RESOURCE2SKILL: Distilling Executable Agent Skills from Human-Created Multimodal Resources 128 UPVOTES · YIJIA FAN ET AL. · MICROSOFT RESEARCH · ARXIV 2606.29538 From Microsoft Research: distilling executable agent skills from human resources and experience, turning procedures into reusable skill libraries instead of hand-written ones.
  3. 3 RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM 128 UPVOTES · MIKHAIL KOMAROV ET AL. · NOVOSIBIRSK STATE UNIVERSITY · ARXIV 2607.11683 A multi-step GraphRAG engine built around a compact, domain-adaptive knowledge graph, addressing the cost and rigidity of existing graph-construction pipelines.
  4. 4 EvolvingWorld: An Open-Schema Framework for Co-Evolving Role-Play Agents and World Model in Interactive Literary World 71 UPVOTES · QING ZONG ET AL. · TENCENT · ARXIV 2607.17250 An open-schema framework and benchmark for character-and-world co-evolution in interactive literary worlds, from Tencent. Characters and their fictional settings change each other over time.
  5. 5 DeepSearch-World: Self-Distillation for Deep Search Agents in a Verifiable Environment 69 UPVOTES · XINYU GENG ET AL. · HKUST · ARXIV 2607.07820 Self-distillation for deep-search agents: instead of fixed teacher-distilled trajectories, agents improve from their own successful search episodes.

2026-07-17 · FRIDAY

  1. 1 VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding 111 UPVOTES · XINHAO LI ET AL. · MULTIMEDIA COMPUTING GROUP-NANJING UNIVERSITY · ARXIV 2607.14935 A fully open video multimodal LLM covering motion, long-video and streaming understanding, positioned as an open reference stack for video assistants.
  2. 2 LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget 97 UPVOTES · CHANGHAI ZHOU ET AL. · MIND LAB · ARXIV 2607.14952 Reinforcement-learning post-training beyond 2 million tokens of context under a fixed GPU budget, attacking the widening gap between inference context lengths and what RL training can reach.
  3. 3 SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning 73 UPVOTES · JINYANG WU ET AL. · ARXIV 2607.14777 Self-evolving on-policy distillation for agentic RL: models trained as interactive agents on long-horizon tasks learn from their own improving policy instead of a fixed teacher.
  4. 4 SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration 52 UPVOTES · YUYAO ZHANG ET AL. · ANT GROUP · ARXIV 2607.15257 Works toward robust open-domain information seeking with tool-integrated LLMs, treating web search as a core model capability rather than a bolted-on tool.
  5. 5 BadWAM: When World-Action Models Dream Right but Act Wrong 37 UPVOTES · QI LI ET AL. · ARXIV 2607.15207 Examines world-action models that predict the future correctly yet still act wrongly - an embodied-control failure mode where dreaming right does not mean acting right.