{
  "schema_version": "paper_public_manifest_v1",
  "paper_id": "alphanist_2026_10",
  "slug": "alphanist",
  "title": "不教步骤，装配技能如何从奖励中涌现？",
  "authors": [],
  "source": {
    "arxiv_id": "2610.13354",
    "pdf_url": "https://arxiv.org/pdf/2610.13354",
    "project_url": "",
    "github_url": "",
    "huggingface_url": "",
    "original_source": "https://arxiv.org/abs/2610.13354"
  },
  "site": {
    "post_url": "/posts/alphanist",
    "canonical_url": "https://haiguangboy.com/posts/alphanist",
    "cover_image": "https://static.haiguangboy.com/papers/alphanist/cover.webp"
  },
  "taxonomy": {
    "domain": "embodied_ai",
    "track": "action_generation",
    "tasks": [
      "embodied_ai",
      "action_generation",
      "robotics",
      "state_prediction",
      "具身智能",
      "机器人",
      "强化学习",
      "机器人装配",
      "alphanist"
    ],
    "related_topics": [
      {
        "paper_id": "an_open_foundation_model_towards_2026_07",
        "title": "An_Open_Foundation_Model_Towards",
        "url": "https://haiguangboy.com/posts/an_open_foundation_model_towards",
        "relation": "contrast",
        "summary": "作者判断异质任务可通过共享终点目标正迁移 contradicts 联合co-training异构人机数据是结构性次优方案",
        "strength": "strong"
      },
      {
        "paper_id": "decompose_and_reorganize_planning_with_primitives_and_visuomotor_policies_learne_2026_08",
        "title": "切换逻辑不该让策略去学",
        "url": "https://haiguangboy.com/posts/dr-lfd",
        "relation": "contrast",
        "summary": "多对装配出现暂存零件后返回的协调行为 contradicts 核心主张:切换逻辑该交给规划器而不是让策略学——这是对「端到端 scaling」的一次结构性反驳",
        "strength": "strong"
      },
      {
        "paper_id": "learning_a_thousand_tasks_in_a_day_2026_08",
        "title": "1 天 1000 任务，靠的是归纳偏置",
        "url": "https://haiguangboy.com/posts/mt3-thousand-tasks",
        "relation": "same_track",
        "summary": "作者判断异质任务可通过共享终点目标正迁移 validates 关键的反向趋势:分解法在 1–10 条时快速起量但 50 条附近见顶,单体 BC 在 10–50 区间加速追赶",
        "strength": "strong"
      },
      {
        "paper_id": "agile_wam_an_agile_tactile_world_action_model_for_contact_rich_robot_control_2026_09",
        "title": "Agile-WAM：轻量视触觉联合预测",
        "url": "https://haiguangboy.com/posts/agile-wam",
        "relation": "same_track",
        "summary": "协调案例不确立通用规划或未见任务泛化 validates 小样本及选优评估不能外推为稳定通用能力",
        "strength": "strong"
      },
      {
        "paper_id": "latent_action_pretraining_through_world_modeling_2026_07",
        "title": "LAWM：为什么动作标签会成为负担",
        "url": "https://haiguangboy.com/posts/latent_action_pretraining_through_world_modeling",
        "relation": "same_track",
        "summary": "LAWM：为什么动作标签会成为负担",
        "strength": "strong"
      },
      {
        "paper_id": "mt_wam_reorienting_the_one_pass_predictive_representation_toward_action_generati_2026_09",
        "title": "MT-WAM：未来监督如何服务动作",
        "url": "https://haiguangboy.com/posts/mt-wam",
        "relation": "same_track",
        "summary": "规模消融同时改变GPU、环境和rollout batch validates 尚未完全分离监督目标、容量和分支结构的因果贡献",
        "strength": "strong"
      }
    ]
  },
  "analyst_take": {
    "type": "author_opinion",
    "text": "💡 多样性如何转化为能力？\n本篇支持的是：共享目标、任务多样性、初态覆盖和经验分配共同作用，能让部分单练难例变得可解。\n\n这还不是“任务越多越聪明”的scaling law。更值得检验的是：扩大任务集合后，正迁移能否持续，以及这些重抓、接触调整和协调行为能否迁移到真机。"
  },
  "ruling": {
    "importance_score": 3.0,
    "one_sentence": "不教步骤，装配技能如何从奖励中涌现？"
  },
  "asset_base_url": "https://static.haiguangboy.com/papers/alphanist",
  "assets": [
    {
      "type": "pdf_screenshot",
      "object_key": "papers/alphanist/page_01.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/page_01.webp",
      "role": "paper_first_page",
      "size_bytes": 228964
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/alphanist/key_figure.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/key_figure.webp",
      "role": "method_figure",
      "size_bytes": 191880
    },
    {
      "type": "cover_image",
      "object_key": "papers/alphanist/cover.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/cover.webp",
      "role": "post_cover",
      "size_bytes": 69152
    },
    {
      "type": "public_brief",
      "object_key": "papers/alphanist/public_brief.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/public_brief.md",
      "role": "public_brief",
      "size_bytes": 3656
    },
    {
      "type": "public_brief",
      "object_key": "papers/alphanist/public_brief.en.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/public_brief.en.md",
      "role": "public_brief_en",
      "size_bytes": 4649
    },
    {
      "type": "public_manifest",
      "object_key": "papers/alphanist/public_manifest.en.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/public_manifest.en.json",
      "role": "public_manifest_en",
      "size_bytes": 6479
    },
    {
      "type": "public_manifest",
      "object_key": "papers/alphanist/public_manifest.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/alphanist/public_manifest.json",
      "role": "public_manifest",
      "size_bytes": 6975
    }
  ],
  "published_at": "2026-10-15T16:19:47+08:00",
  "created_at": "2026-10-15T16:19:47+08:00",
  "updated_at": "2026-10-15T16:19:59+08:00"
}