{
  "title": "Daily Research Brief 2026-04-25",
  "url": "/en/posts/research-brief-2026-04-25/",
  "permalink": "https://hackcv.com/en/posts/research-brief-2026-04-25/",
  "date": "2026-04-25",
  "lastmod": "2026-04-25",
  "author": "",
  "description": "Daily research brief — AI / LLM / Agent / Computer Vision / Audio-Video / Engineering",
  "categories": ["Research Brief"],
  "tags": ["AI","LLM","Agent","Computer Vision","Audio-Video","Engineering","Daily Brief"],
  "cover": "https://picsum.photos/seed/daily-research-brief-2026-04-25/1200/675",
  "readingTime": 1,
  "wordCount": 142,
  "content": "\u003ch1 id=\"daily-research-brief-2026-04-25\"\u003eDaily Research Brief 2026-04-25\u003c/h1\u003e\n\u003cp\u003e📊 Token usage: estimated from retrieval and writing scale.\u003c/p\u003e\n\u003cp\u003eCovers the latest AI research, open source and industry moves, updated daily.\u003c/p\u003e\n\u003chr\u003e\n\u003ch2 id=\"editors-note\"\u003eEditor\u0026rsquo;s Note\u003c/h2\u003e\n\u003cp\u003eMeta\u0026rsquo;s loss is Thinking Machines\u0026rsquo; gain; ComfyUI hit a $500M valuation as creators seek more control; Google plans to invest up to $40B in Anthropic in cash and compute.\u003c/p\u003e\n\u003ch2 id=\"1-latest-arxiv-papers\"\u003e1. Latest arXiv Papers\u003c/h2\u003e\n\u003col\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eSeeing Fast and Slow: Learning the Flow of Time in Videos\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21931\"\u003ehttps://arxiv.org/abs/2604.21931\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eSeeing Without Eyes: 4D Human-Scene Understanding from Audio\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21926\"\u003ehttps://arxiv.org/abs/2604.21926\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eTemporal Taskification in Streaming Continual Learning\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21930\"\u003ehttps://arxiv.org/abs/2604.21930\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eFine-Tuning Regimes Define Distinct Continual Learning Behaviors\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21927\"\u003ehttps://arxiv.org/abs/2604.21927\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eWhen Prompts Override Vision: Prompt-Induced Hallucination in VLMs\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21911\"\u003ehttps://arxiv.org/abs/2604.21911\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eFrom Research Question to Scientific Workflow: LLM-Driven Experiment Design\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21910\"\u003ehttps://arxiv.org/abs/2604.21910\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eEvaluation of Automatic Speech Recognition Using Diverse Acoustic Conditions\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21928\"\u003ehttps://arxiv.org/abs/2604.21928\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eMathDuels: Evaluating LLMs as Problem Posers and Solvers\u003c/strong\u003e — \u003ca href=\"https://arxiv.org/abs/2604.21916\"\u003ehttps://arxiv.org/abs/2604.21916\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n",
  "summary": "Daily Research Brief 2026-04-25 📊 Token usage: estimated from retrieval and writing scale.\nCovers the latest AI research, open source and industry moves, updated daily.\nEditor\u0026rsquo;s Note Meta\u0026rsquo;s loss is Thinking Machines\u0026rsquo; gain; ComfyUI hit a $500M valuation as creators seek more control; Google plans to invest up to $40B in Anthropic in cash and compute.\n1. Latest arXiv Papers Seeing Fast and Slow: Learning the Flow of Time in Videos — https://arxiv.org/abs/2604.21931\n"
}