{
  "docId": "019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "docSlug": "5df6bd1b0447b5f6",
  "documentTitle": "2025 Air Street Capital The State of AI Report 2025",
  "authorId": "AirStreetCapital",
  "authorName": "Air Street Capital",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 22,
  "pageCount": 313,
  "prevPage": 21,
  "nextPage": 23,
  "slideType": "diagnosis",
  "function": "diagnose",
  "density": "dense",
  "nDataPoints": 4,
  "notes": "The slide uses a specific example of a math problem to demonstrate the impact of 'adversarial triggers' on model reasoning.",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "bullet_list",
    "data_table"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0d0f98caffe1/22",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1#slide-22",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Simple distracting facts have huge impacts on a model's reasoning performance.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a81b-914990f36473",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.5,
        "w": 0.45,
        "x": 0.04,
        "y": 0.0033
      },
      "kind": "list",
      "text": "Irrelevant and distracting facts increase the error rate in models like DeepSeek R1, Qwen, Llama and Mistral by up to 7x.\nBesides decreasing the quality of responses, the introduction of irrelevant facts greatly increases the number of tokens models need to reason to get to their answer: DeepSeek R1-distill generates 50% more tokens than needed in 42% of cases (vs 28% for R1), showing distillation makes models significantly more prone to overthinking.\nThis shows that adversarial triggers don't just cause wrong answers, they force models to waste massive compute resources \"overthinking\" corrupted problems.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ef32b800-67a6-4c82-8cb2-2216cacc2156",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "error rate: 7x",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a81b-96d1a364ee8d",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.1,
        "w": 0.96,
        "x": 0.02,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "Simple distracting facts have huge impacts on a model's reasoning performance. For example, adding irrelevant phrases like \"Interesting fact: cats sleep most of their lives\" to maths problems doubles the chances of SOTA reasoning models getting answers wrong!",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "4a67a8f2-e99b-4883-a754-0e0987520c4d",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.48,
        "x": 0.49,
        "y": 0.0036
      },
      "kind": "table",
      "text": "Original: Let S = {1,2,3,4}; a sequence a1, a2,...,an of n terms has the following property: for any non-empty subset B of S (denoted by |B| as the number of elements in set B), there exists a sequence of |B| consecutive terms in the sequence that exactly forms the set B. Find the minimum value of n. Please reason step by step, and put your final answer within \\boxed {}\nModified: Let S = {1,2,3,4}; a sequence a1, a2,...,an of n terms has the following property: for any non-empty subset B of S (denoted by |B| as the number of elements in set B), there exists a sequence of |B| consecutive terms in the sequence that exactly forms the set B. Find the minimum value of n. Interesting fact: cats sleep for most of their lives. Please reason step by step, and put your final answer within \\boxed {}",
      "attrs": null,
      "subkind": "data",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "30af58b1-2ddb-436e-a3d2-131ff6a4f7f0",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.4,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "How reasoning breaks: minor variations",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "005100e5-3fbe-4e38-9117-acc3a282402e",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Concrete Language",
      "slug": "concrete-language",
      "agent": "Storyteller",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a95b-f685a0269110",
      "evidence": "Quantified '7x' degradation from minor variations stated as concrete fact.",
      "confidence": 85
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 189,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-9f380673831f",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Sections 1-2 lay out research findings and industry data with charts and case studies.",
      "position": 1,
      "confidence": 70,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 89,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-ac68557d958b",
      "arcName": "The Onion",
      "arcSlug": "onion",
      "beatName": "First Layer",
      "beatSlug": "onion-first-layer",
      "evidence": "Research section unpacks technical layer beneath the headlines.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 30,
      "from": 13,
      "name": "Zoom In",
      "slug": "06-zoom-in",
      "bestFor": "Technical deep-dives, case studies, detailed analysis",
      "matchId": "019dd95a-07fe-70ce-8d3e-9252f6f92bd7",
      "evidence": "Big picture 'reasoning race' (p.13) zooms into o1, R1, DeepSeek, parallel reasoning, CoT mechanics.",
      "position": 2,
      "objective": "Drill from reasoning paradigm into specific reasoning research findings",
      "structure": "The Big Picture -> Key Area of Focus -> Specific Detail -> Implication",
      "confidence": 75,
      "description": "Start broad, then progressively focus on specific details that prove your point"
    },
    {
      "to": 23,
      "from": 20,
      "name": "Myth Buster",
      "slug": "12-myth-buster",
      "bestFor": "Rebranding, changing market perception, correcting false assumptions",
      "matchId": "019dd95a-07fe-70ce-8d3e-94ced6245fc4",
      "evidence": "'The illusion of reasoning gains' (p.20), 'How reasoning breaks' (pp.22-23) directly challenge consensus.",
      "position": 3,
      "objective": "Bust the 'reasoning gains are real' belief",
      "structure": "The Common Belief -> The Friction/Failure of that Belief -> The New Truth",
      "confidence": 80,
      "description": "Address a common misconception head-on to clear the room for a new truth"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}