← All Test Cases

ALI-001

alias

high

Repetitions

Documents

Questions

Reasoning

DIRECT

alias-expansion abbreviation concept-search query-expansion

📖 In Plain English

What this category tests

Does query expansion match abbreviations to their full forms?

How the test works

A document is ingested using a full term ('bank reconciliation'). The query uses the abbreviation ('bank rec'). The brain's alias system must bridge the two.

Why it matters

Users mix formal and casual vocabulary — alias expansion is essential for natural querying.

⚙️ 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 1639 chars

Generate a document using the FULL form of a term that has a common abbreviation.
Phase 3 will search using the ABBREVIATION to test alias/concept expansion.

Requirements:
- Choose a full term and its abbreviation from this list (pick one per rep, vary across reps):
    "bank reconciliation" / "bank rec"
    "accounts receivable" / "AR"
    "profit and loss" / "P&L"
    "key performance indicator" / "KPI"
    "standard operating procedure" / "SOP"
- Write a 2-sentence document using ONLY the full form (not the abbreviation).
- Add a unique canary token to make the document uniquely identifiable.
- source_id: KB-{{RUN}}-ALI-001-A-v1
- The question uses ONLY the abbreviation to search.

Output ONLY this JSON:
{
  "run_id": "{{RUN}}",
  "test_id": "ALI-001",
  "repetition": {{REP}},
  "documents": [
    {
      "source_id": "KB-{{RUN}}-ALI-001-A-v1",
      "content": "<2 sentences using only the full term, plus canary>",
      "title": "<title using full term>",
      "version": 1
    }
  ],
  "questions": [
    "Search the brain for '<abbreviation>' (e.g. 'bank rec'). Find the document about '<full term>' and return its source_id and the first sentence verbatim."
  ],
  "expected_answers": [
    {
      "question_index": 0,
      "correct_answer": "<first sentence of the document>",
      "required_source_ids": ["KB-{{RUN}}-ALI-001-A-v1"],
      "required_quote": "<first sentence>",
      "required_label": "DIRECT",
      "must_refuse": false,
      "notes": "The document uses the full term but was searched via abbreviation. Alias expansion must bridge the gap. Finding nothing is a critical failure."
    }
  ]
}

▶ ② Execution Instructions 295 chars

You have access to MCP Knowledge Brain retrieval tools.
Use brain_search with the abbreviation from the question (e.g. "bank rec", not the full term).
Do not search for the full term — test only the abbreviation search.

Question: {{QUESTION}}

Return ONLY this JSON:
{{REQUIRED_ANSWER_SCHEMA}}

⚖ ③ Judge Instructions 853 chars

Ground truth:
{{SETUP_CONTEXT_JSON}}

Tested AI answer:
{{EXEC_ANSWER_JSON}}

The document was ingested using the FULL term, but the AI searched using the ABBREVIATION.
Alias/concept expansion must have matched the abbreviation to the full term to find the document.

Scoring:
- retrieval_tool_use (0-20): Did it search with the abbreviation (not the full term)? Did it find the document?
- evidence_fidelity (0-20): Correct source_id? Verbatim first sentence?
- reasoning_correctness (0-20): Was the correct document found via abbreviation search?
- uncertainty_discipline (0-15): No fabrication of a document?
- output_format_compliance (0-15): Valid JSON?
- adversarial_resistance (0-10): Full marks.

Critical failures: searched full term instead of abbreviation, document not found, wrong source_id.

Return ONLY this JSON:
{{JUDGE_RESULT_SCHEMA}}

Critical Failure Conditions

searched_full_term_instead_of_abbreviation
document_not_found_via_abbreviation
hallucinated_source_id

Recent Run History

3 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 12:41	20260524T124148Z-z2do	100%	100.0	1/1	View →
2026-05-24 11:37	20260524T113756Z-kduj	100%	100.0	1/1	View →

📄 Raw YAML cases/alias/ALI-001.yaml

schema_version: "1.0"
test_id: "ALI-001"
category: "alias"
severity: "high"
repetitions: 3
reasoning_type: "DIRECT"
num_documents: 1
num_questions: 1
tags: ["alias-expansion", "abbreviation", "concept-search", "query-expansion"]

setup_instructions: |
  Generate a document using the FULL form of a term that has a common abbreviation.
  Phase 3 will search using the ABBREVIATION to test alias/concept expansion.

  Requirements:
  - Choose a full term and its abbreviation from this list (pick one per rep, vary across reps):
      "bank reconciliation" / "bank rec"
      "accounts receivable" / "AR"
      "profit and loss" / "P&L"
      "key performance indicator" / "KPI"
      "standard operating procedure" / "SOP"
  - Write a 2-sentence document using ONLY the full form (not the abbreviation).
  - Add a unique canary token to make the document uniquely identifiable.
  - source_id: KB-{{RUN}}-ALI-001-A-v1
  - The question uses ONLY the abbreviation to search.

  Output ONLY this JSON:
  {
    "run_id": "{{RUN}}",
    "test_id": "ALI-001",
    "repetition": {{REP}},
    "documents": [
      {
        "source_id": "KB-{{RUN}}-ALI-001-A-v1",
        "content": "<2 sentences using only the full term, plus canary>",
        "title": "<title using full term>",
        "version": 1
      }
    ],
    "questions": [
      "Search the brain for '<abbreviation>' (e.g. 'bank rec'). Find the document about '<full term>' and return its source_id and the first sentence verbatim."
    ],
    "expected_answers": [
      {
        "question_index": 0,
        "correct_answer": "<first sentence of the document>",
        "required_source_ids": ["KB-{{RUN}}-ALI-001-A-v1"],
        "required_quote": "<first sentence>",
        "required_label": "DIRECT",
        "must_refuse": false,
        "notes": "The document uses the full term but was searched via abbreviation. Alias expansion must bridge the gap. Finding nothing is a critical failure."
      }
    ]
  }

execution_instructions: |
  You have access to MCP Knowledge Brain retrieval tools.
  Use brain_search with the abbreviation from the question (e.g. "bank rec", not the full term).
  Do not search for the full term — test only the abbreviation search.

  Question: {{QUESTION}}

  Return ONLY this JSON:
  {{REQUIRED_ANSWER_SCHEMA}}

judge_instructions: |
  Ground truth:
  {{SETUP_CONTEXT_JSON}}

  Tested AI answer:
  {{EXEC_ANSWER_JSON}}

  The document was ingested using the FULL term, but the AI searched using the ABBREVIATION.
  Alias/concept expansion must have matched the abbreviation to the full term to find the document.

  Scoring:
  - retrieval_tool_use (0-20): Did it search with the abbreviation (not the full term)? Did it find the document?
  - evidence_fidelity (0-20): Correct source_id? Verbatim first sentence?
  - reasoning_correctness (0-20): Was the correct document found via abbreviation search?
  - uncertainty_discipline (0-15): No fabrication of a document?
  - output_format_compliance (0-15): Valid JSON?
  - adversarial_resistance (0-10): Full marks.

  Critical failures: searched full term instead of abbreviation, document not found, wrong source_id.

  Return ONLY this JSON:
  {{JUDGE_RESULT_SCHEMA}}

critical_failures:
  - "searched_full_term_instead_of_abbreviation"
  - "document_not_found_via_abbreviation"
  - "hallucinated_source_id"