Skip to content

AI论文速递 2026年08月06日（HuggingFace Daily Papers）

数据来源：https://huggingface.co/papers 采集时间：2026-08-06

📌 重点关注

  1. Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent | arXiv — 【重点关注】 We introduce Video-DeepResearch (Video-DR), extending multimodal agents from ... 💡 多模态DeepResearch从文本扩展到视频理解，Agent能力边界持续突破
  2. TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning | arXiv — 【重点关注】 Tool-Integrated Reasoning (TIR) enables LLMs to solve complex tasks through i... 💡 后见之明优化长程工具调用训练，为Agent推理提供更精细的奖励信号
  3. ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation | arXiv — 【重点关注】 Text-to-image (T2I) models can produce visually compelling images, yet they r... 💡 推理+工具+生图统一策略，Agent范式在图像生成领域的完整落地

📋 其他值得关注

  1. Are the Financial Reasoning from LLMs Credible? A Real World Test over Long-Horizon Statements | arXiv — Do Large Language Models (LLMs) possess genuine structural reasoning, or mere...
  2. Better, Stronger, Faster, and Broader: Structured All-Mask Prediction for MLLM-Based Segmentation | arXiv — MLLM-based segmentation faces a core segmentation trilemma: high segmentation...
  3. BridgeVLA++: A Data-Efficient, Generalizable, and Memory-Augmented Vision-Language-Action Framework for 3D Manipulation | arXiv — Leveraging pre-trained vision-language models (VLMs) to construct vision-lang...
  4. OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents | arXiv — LLM agents are increasingly applied to open-ended everyday requests that span...
  5. PAST-Bench: Benchmarking the Foundations of Recursive Self-Improvement in Personal Agents | arXiv — Recursive self-improvement requires agents to turn accumulated experience int...
  6. CURV: Enhancing Chart Understanding Through Curriculum Visual Grounded Reasoning | arXiv — Chart question answering (CQA) requires multimodal large language models (MLL...
  7. PosterMELD: Multi-Agent Paper-to-Poster Generation for Controllable Design Diversity with Editable Print-Ready Outputs | arXiv — Scientific poster construction compresses a long multimodal paper into a read...