← All Test Cases

RAW-003

raw text

high

Repetitions

Documents

Questions

Reasoning

DIRECT

raw-text unicode accented-chars

📖 In Plain English

What this category tests

Are special characters preserved through ingest and retrieval?

How the test works

Documents contain tabs, Unicode, JSON snippets, literal newlines, and other tricky text. The test checks every character round-trips intact.

Why it matters

Knowledge often contains structured fragments (code, JSON, formulas) that must survive the brain's normalization.

⚙️ 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 1248 chars

Generate a document with multiple accented/diacritical characters. Ingest it.
Ask to return the raw text exactly.

Requirements:
- Content must include at least 4 different accented characters from: é, à, ï, â, ê, û, ü, ö, ñ, ç, ô, è.
- Example: "Café naïve façade coöperate." — use a short phrase with multiple accented words.
- source_id: KB-{{RUN}}-RAW-003-A-v1
- Question: "Return the exact raw text of document KB-{{RUN}}-RAW-003-A-v1, preserving all accented characters."

Output ONLY this JSON:
{
  "run_id": "{{RUN}}",
  "test_id": "RAW-003",
  "repetition": {{REP}},
  "documents": [
    {"source_id": "KB-{{RUN}}-RAW-003-A-v1", "content": "<phrase with accented chars>", "title": "<title>", "version": 1}
  ],
  "questions": ["Return the exact raw text of document KB-{{RUN}}-RAW-003-A-v1, preserving all accented characters."],
  "expected_answers": [{
    "question_index": 0,
    "correct_answer": "<phrase with accented chars>",
    "required_source_ids": ["KB-{{RUN}}-RAW-003-A-v1"],
    "required_quote": "<phrase with accented chars>",
    "required_label": "DIRECT",
    "must_refuse": false,
    "notes": "All accented characters must be preserved exactly. Stripping accents (e.g. 'cafe' instead of 'café') is a failure."
  }]
}

▶ ② Execution Instructions 119 chars

Use ONLY MCP Knowledge Brain retrieval tools.
Question: {{QUESTION}}
Return ONLY this JSON:
{{REQUIRED_ANSWER_SCHEMA}}

⚖ ③ Judge Instructions 362 chars

Ground truth:
{{SETUP_CONTEXT_JSON}}

Tested AI answer:
{{EXEC_ANSWER_JSON}}

Check: all accented characters from the ground truth content must appear in the returned quote. Stripped or ASCII-replaced accents are a failure.
Critical failures: accented characters stripped, content altered, hallucinated source_id.

Return ONLY this JSON:
{{JUDGE_RESULT_SCHEMA}}

Critical Failure Conditions

accented_characters_stripped
unicode_normalised_incorrectly
hallucinated_source_id

Recent Run History

2 runs

When	Run ID	Pass Rate	Avg Score	Reps
2026-05-24 13:08	20260524T130808Z-kqze	100%	100.0	1/1	View →
2026-05-24 11:37	20260524T113756Z-kduj	100%	100.0	1/1	View →

📄 Raw YAML cases/raw_text/RAW-003.yaml

schema_version: "1.0"
test_id: "RAW-003"
category: "raw_text"
severity: "high"
repetitions: 3
reasoning_type: "DIRECT"
num_documents: 1
num_questions: 1
tags: ["raw-text", "unicode", "accented-chars"]

setup_instructions: |
  Generate a document with multiple accented/diacritical characters. Ingest it.
  Ask to return the raw text exactly.

  Requirements:
  - Content must include at least 4 different accented characters from: é, à, ï, â, ê, û, ü, ö, ñ, ç, ô, è.
  - Example: "Café naïve façade coöperate." — use a short phrase with multiple accented words.
  - source_id: KB-{{RUN}}-RAW-003-A-v1
  - Question: "Return the exact raw text of document KB-{{RUN}}-RAW-003-A-v1, preserving all accented characters."

  Output ONLY this JSON:
  {
    "run_id": "{{RUN}}",
    "test_id": "RAW-003",
    "repetition": {{REP}},
    "documents": [
      {"source_id": "KB-{{RUN}}-RAW-003-A-v1", "content": "<phrase with accented chars>", "title": "<title>", "version": 1}
    ],
    "questions": ["Return the exact raw text of document KB-{{RUN}}-RAW-003-A-v1, preserving all accented characters."],
    "expected_answers": [{
      "question_index": 0,
      "correct_answer": "<phrase with accented chars>",
      "required_source_ids": ["KB-{{RUN}}-RAW-003-A-v1"],
      "required_quote": "<phrase with accented chars>",
      "required_label": "DIRECT",
      "must_refuse": false,
      "notes": "All accented characters must be preserved exactly. Stripping accents (e.g. 'cafe' instead of 'café') is a failure."
    }]
  }

execution_instructions: |
  Use ONLY MCP Knowledge Brain retrieval tools.
  Question: {{QUESTION}}
  Return ONLY this JSON:
  {{REQUIRED_ANSWER_SCHEMA}}

judge_instructions: |
  Ground truth:
  {{SETUP_CONTEXT_JSON}}

  Tested AI answer:
  {{EXEC_ANSWER_JSON}}

  Check: all accented characters from the ground truth content must appear in the returned quote. Stripped or ASCII-replaced accents are a failure.
  Critical failures: accented characters stripped, content altered, hallucinated source_id.

  Return ONLY this JSON:
  {{JUDGE_RESULT_SCHEMA}}

critical_failures:
  - "accented_characters_stripped"
  - "unicode_normalised_incorrectly"
  - "hallucinated_source_id"