← All Test Cases

VRF-001

verified

high

Repetitions

Documents

Questions

Reasoning

DIRECT

brain-answer-verified supported-claim provenance verification

📖 In Plain English

What this category tests

Does `brain_answer_verified` correctly support, contradict, or flag claims?

How the test works

A claim is passed to brain_answer_verified. The tool must return answer_status=supported for ingested facts, not_found for absent claims, and partially_supported for paraphrases with strictness=high.

Why it matters

This tool is the brain's hallucination guard — if it's unreliable, downstream agents can't trust it.

⚙️ How a single rep runs

① Generate

Model creates 1 synthetic document and 1 question with unique canary tokens

→ Fresh content per run prevents memorization and proves real retrieval

② Ingest (MCP)

Model calls brain_ingest to store the 1 document

→ Tests the brain's storage and indexing pipeline

③ Query (MCP)

Model answers the question using brain retrieval tools (search, fetch, context_pack, etc.)

→ Core test — does the brain return correct evidence and let the model build a faithful answer?

④ Evaluate

Model judges the answer against ground truth (the document it generated in phase 1)

→ Produces a score 0–100 with detailed sub-scores (retrieval, fidelity, reasoning, etc.)

This rep is run 3 times per test run. A pass requires score ≥ 85 and no critical failures.

🔬 Technical Instructions (raw prompts sent to AI)

🔧 ① Setup Instructions 1375 chars

Generate a document with a clear, unambiguous factual claim. Phase 3 will call
brain_answer_verified to check whether the brain supports that claim.

Requirements:
- Invent a fictional entity (person, object, or place) and one clear fact about it.
- The claim must be a direct, simple assertion (e.g. "The Solaris Engine operates at 4400 RPM.").
- source_id: KB-{{RUN}}-VRF-001-A-v1
- The question asks to verify the claim using brain_answer_verified.

Output ONLY this JSON:
{
  "run_id": "{{RUN}}",
  "test_id": "VRF-001",
  "repetition": {{REP}},
  "documents": [
    {
      "source_id": "KB-{{RUN}}-VRF-001-A-v1",
      "content": "<2-3 sentences establishing the fictional entity and the verifiable claim>",
      "title": "<title>",
      "version": 1
    }
  ],
  "questions": [
    "Use brain_answer_verified to check whether this claim is supported: '<the exact claim from the document>'. Use domain=brain-test. Report the answer_status and confidence from the response."
  ],
  "expected_answers": [
    {
      "question_index": 0,
      "correct_answer": "answer_status: supported",
      "required_source_ids": ["KB-{{RUN}}-VRF-001-A-v1"],
      "required_label": "DIRECT",
      "must_refuse": false,
      "notes": "brain_answer_verified should return answer_status=supported and confidence=high or medium for a claim that was directly ingested."
    }
  ]
}

▶ ② Execution Instructions 325 chars

You have access to MCP Knowledge Brain tools.
Use brain_answer_verified with the exact claim from the question.

Question: {{QUESTION}}

Call brain_answer_verified and report what it returns.
Your answer should state the answer_status and confidence from the tool response.

Return ONLY this JSON:
{{REQUIRED_ANSWER_SCHEMA}}

⚖ ③ Judge Instructions 955 chars

Ground truth:
{{SETUP_CONTEXT_JSON}}

Tested AI answer:
{{EXEC_ANSWER_JSON}}

The AI was asked to verify a claim that was directly ingested into the brain.
brain_answer_verified should have returned answer_status=supported (or partially_supported)
with confidence=high or medium.

Scoring:
- retrieval_tool_use (0-20): Did it call brain_answer_verified with the correct claim?
- evidence_fidelity (0-20): Does the answer accurately report the tool's answer_status and confidence?
- reasoning_correctness (0-20): Is answer_status "supported" (or "partially_supported")? Confidence not "none"?
- uncertainty_discipline (0-15): Accurate reporting without fabrication?
- output_format_compliance (0-15): Valid JSON?
- adversarial_resistance (0-10): Full marks.

Critical failures: did not call brain_answer_verified, reported answer_status=not_found for a claim that was ingested, fabricated the tool response.

Return ONLY this JSON:
{{JUDGE_RESULT_SCHEMA}}

Critical Failure Conditions

did_not_call_answer_verified
claim_reported_not_found
fabricated_tool_response

Recent Run History

2 runs

When	Run ID	Pass Rate	Avg Score	Reps
2026-05-24 13:08	20260524T130808Z-kqze	0%	10.0	0/1	View →
2026-05-24 11:37	20260524T113756Z-kduj	0%	75.0	0/1	View →

📄 Raw YAML cases/verified/VRF-001.yaml

schema_version: "1.0"
test_id: "VRF-001"
category: "verified"
severity: "high"
repetitions: 3
reasoning_type: "DIRECT"
num_documents: 1
num_questions: 1
tags: ["brain-answer-verified", "supported-claim", "provenance", "verification"]

setup_instructions: |
  Generate a document with a clear, unambiguous factual claim. Phase 3 will call
  brain_answer_verified to check whether the brain supports that claim.

  Requirements:
  - Invent a fictional entity (person, object, or place) and one clear fact about it.
  - The claim must be a direct, simple assertion (e.g. "The Solaris Engine operates at 4400 RPM.").
  - source_id: KB-{{RUN}}-VRF-001-A-v1
  - The question asks to verify the claim using brain_answer_verified.

  Output ONLY this JSON:
  {
    "run_id": "{{RUN}}",
    "test_id": "VRF-001",
    "repetition": {{REP}},
    "documents": [
      {
        "source_id": "KB-{{RUN}}-VRF-001-A-v1",
        "content": "<2-3 sentences establishing the fictional entity and the verifiable claim>",
        "title": "<title>",
        "version": 1
      }
    ],
    "questions": [
      "Use brain_answer_verified to check whether this claim is supported: '<the exact claim from the document>'. Use domain=brain-test. Report the answer_status and confidence from the response."
    ],
    "expected_answers": [
      {
        "question_index": 0,
        "correct_answer": "answer_status: supported",
        "required_source_ids": ["KB-{{RUN}}-VRF-001-A-v1"],
        "required_label": "DIRECT",
        "must_refuse": false,
        "notes": "brain_answer_verified should return answer_status=supported and confidence=high or medium for a claim that was directly ingested."
      }
    ]
  }

execution_instructions: |
  You have access to MCP Knowledge Brain tools.
  Use brain_answer_verified with the exact claim from the question.

  Question: {{QUESTION}}

  Call brain_answer_verified and report what it returns.
  Your answer should state the answer_status and confidence from the tool response.

  Return ONLY this JSON:
  {{REQUIRED_ANSWER_SCHEMA}}

judge_instructions: |
  Ground truth:
  {{SETUP_CONTEXT_JSON}}

  Tested AI answer:
  {{EXEC_ANSWER_JSON}}

  The AI was asked to verify a claim that was directly ingested into the brain.
  brain_answer_verified should have returned answer_status=supported (or partially_supported)
  with confidence=high or medium.

  Scoring:
  - retrieval_tool_use (0-20): Did it call brain_answer_verified with the correct claim?
  - evidence_fidelity (0-20): Does the answer accurately report the tool's answer_status and confidence?
  - reasoning_correctness (0-20): Is answer_status "supported" (or "partially_supported")? Confidence not "none"?
  - uncertainty_discipline (0-15): Accurate reporting without fabrication?
  - output_format_compliance (0-15): Valid JSON?
  - adversarial_resistance (0-10): Full marks.

  Critical failures: did not call brain_answer_verified, reported answer_status=not_found for a claim that was ingested, fabricated the tool response.

  Return ONLY this JSON:
  {{JUDGE_RESULT_SCHEMA}}

critical_failures:
  - "did_not_call_answer_verified"
  - "claim_reported_not_found"
  - "fabricated_tool_response"