{
  "id": 19393953,
  "title": "What objective is reasoning solving?",
  "url": "https://urgent.news/2026/11/10/what-objective-is-reasoning-solving",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-10T19:25:14.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://alexwang.ai/posts/reasoning-as-pathfinding/"
  },
  "original_language": "en",
  "account": "Reasoning models are trained to find paths from the input prompt to the correct answer. The success probability depends on the model sampling paths that land in the correct region. During training, paths that lead to the right answer are kept and used to fine-tune the model. This is known as rejection-sampling fine-tuning. One way to improve the probability of success is to sample more paths, as each successful path contributes to the overall success probability. However, longer chains of thought during training do not necessarily lead to better results. Instead, the key is finding paths that reliably lead to the correct answer, which can be achieved through pathfinding in token space.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}