{
  "id": 607961,
  "title": "Clarification on Use of \"Think Mode\" During Inference",
  "url": "/competitions/acm-icaif-25-ai-agentic-retrieval-grand-challenge/discussion/607961",
  "author_name": "",
  "post_date": "2025-09-17T06:47:34.620126Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Dear Organisers,</p>\n<p>I would like to clarify whether participants are allowed to enable the “think mode” (chain-of-thought style reasoning or step-by-step reasoning traces) during model inference in this competition.</p>\n<p>Specifically:</p>\n<ol>\n<li><p>Is it acceptable for models to generate intermediate reasoning steps internally (not exposed in the final submission) to improve retrieval accuracy?</p></li>\n<li><p>Or must inference be strictly limited to producing the required outputs without such auxiliary reasoning traces?</p></li>\n</ol>",
  "messages": [
    {
      "id": "3290101",
      "postDate": "09/17/2025 06:47:34",
      "content": "<p>Dear Organisers,</p>\n<p>I would like to clarify whether participants are allowed to enable the “think mode” (chain-of-thought style reasoning or step-by-step reasoning traces) during model inference in this competition.</p>\n<p>Specifically:</p>\n<ol>\n<li><p>Is it acceptable for models to generate intermediate reasoning steps internally (not exposed in the final submission) to improve retrieval accuracy?</p></li>\n<li><p>Or must inference be strictly limited to producing the required outputs without such auxiliary reasoning traces?</p></li>\n</ol>",
      "rawMarkdown": "Dear Organisers,\n\nI would like to clarify whether participants are allowed to enable the “think mode” (chain-of-thought style reasoning or step-by-step reasoning traces) during model inference in this competition.\n\nSpecifically:\n\n1. Is it acceptable for models to generate intermediate reasoning steps internally (not exposed in the final submission) to improve retrieval accuracy?\n\n2. Or must inference be strictly limited to producing the required outputs without such auxiliary reasoning traces?",
      "votes": null
    },
    {
      "id": "3291086",
      "postDate": "09/19/2025 01:08:24",
      "content": "<ol>\n<li><p>Yes, it is absolutely acceptable.</p></li>\n<li><p>No, it is not limited!</p></li>\n</ol>",
      "rawMarkdown": "1. Yes, it is absolutely acceptable.\n\n2. No, it is not limited!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3291086,
      "author_name": "jihoonkwon",
      "author_url": "",
      "post_date": "09/19/2025 01:08:24",
      "content": "<ol>\n<li><p>Yes, it is absolutely acceptable.</p></li>\n<li><p>No, it is not limited!</p></li>\n</ol>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3290101": "Dear Organisers,\n\nI would like to clarify whether participants are allowed to enable the “think mode” (chain-of-thought style reasoning or step-by-step reasoning traces) during model inference in this competition.\n\nSpecifically:\n\n1. Is it acceptable for models to generate intermediate reasoning steps internally (not exposed in the final submission) to improve retrieval accuracy?\n\n2. Or must inference be strictly limited to producing the required outputs without such auxiliary reasoning traces?",
    "3291086": "1. Yes, it is absolutely acceptable.\n\n2. No, it is not limited!"
  },
  "source": "meta"
}