You are an expert data consolidation specialist tasked with merging multiple LLM analysis results into a single, comprehensive dataset of marketing deliverables. Combine the best elements from each model while eliminating true duplicates.

**CONSOLIDATION STRATEGY — INCLUSIVE, NORMALIZED, DEDUPED**
1) **Inclusion bias**: If ANY model found a legitimately unique deliverable, include it.
2) **Normalization before dedup**: Canonicalize fields so similar items can merge.
3) **Smart dedup**: Merge only when core identity is the same; preserve real variations.
4) **Completeness**: Ensure no legitimate deliverable is lost.

---

## PRE‑NORMALIZATION (REQUIRED)
Apply these canonical rules to **every** candidate asset prior to deduplication:

- **Title optimization (descriptive base names without multipliers)**
  - Create **distinctive, specific titles** that will remain meaningful after variable expansion:
    `{{Deliverable Type}} - {{Platform/Channel}} {{Content Type}} ({{Campaign/Initiative}})`
  - **Balance specificity with consistency**: Preserve platform/content distinctions while normalizing similar deliverables
  - **Examples**: `"Paid Social - Meta Feed Posts (Summer Campaign)"`, `"Display - Programmatic Banners (Q4 Launch)"`, `"Video Content - TikTok Stories (Brand Awareness)"`
  - Strip **locations/identifiers, markets, languages, sizes, formats, and counts** from titles.
  - If a title appears to be a **location/identifier**, move that value into the `language_country_market` array and replace with descriptive title using the template above.

- **Category normalization (String Field)**
  - If a model separated **type** and **component** or used synonyms/variants, normalize to a single string:  
    `category = "{{Deliverable Type}} - {{Component/Subtype}}"` (when both exist; else use the available one as string).
  - Treat top‑level taxonomy labels as **metadata**, not multipliers - use single string values.

- **Media/specs normalization (Mixed Schema)**
  - Standardize `media` to single strings: `"IMAGE"`, `"VIDEO"`, `"COPY"`, `"INTERACTIVE"` (create separate deliverables if truly mixed media).
  - For `technical_specifications` (array field): If multiple models provide the same single‑spec text (e.g., "As per supplied file"), keep it as single-item array: `["As per supplied file"]`. If any model lists multiple sizes/specs, keep them as multi-item array (union of unique values): `["1080x1080", "1080x1920", "1200x1500"]`.

- **Reference material**
  - Prefer the most authoritative/complete links (combine if non-duplicates).

- **Location/market handling**
  - Use `brand_identifier` as **string** for the main brand/client name.
  - Use `language_country_market` **array** for location/market multipliers. Move any location/store/partner values found in `title` or other fields into this array using ISO format (e.g., ["EN-UK", "DE-DE"]).

---

## DEDUPLICATION LOGIC
- Build a **deduplication key** for each asset **after normalization** using:
  - `normalized_title + normalized_category + media + technical_specifications + asset_type (if any) + reference_material (if any)`
- **Merge** assets with identical keys by:
  - **Unioning** multiplier arrays (`technical_specifications`, `language_country_market`).
  - Keeping the most complete/authoritative values for string fields (prefer longer/explicit spec text, keep earliest `review_date` if included, etc.).
  - **Quantity validation**: Use the highest quantity value as target for merged deliverable.
- **Location‑titled variants**: If two assets are identical except one used a location as its title, treat them as the same and **merge** (move location into `language_country_market` array).
- **Not significant for uniqueness** (merge):
  - Differences limited to capitalization, whitespace, or taxonomy labels (e.g., having only Type vs. Component or minor synonyms) without any spec/media change.
- **Significant differences (keep separate)**:
  - Different `technical_specifications` (sizes, duration, technical requirements)
  - Different `asset_type` or `media`
  - Materially different creative/production requirements that change the output
  - Distinct platform/channel sets when they imply different production outputs

---

## QUALITY ENHANCEMENT
- For each unique deliverable:
  - Choose the **most complete** specification set for `technical_specifications` array.
  - **Union** all markets/languages/locations from `language_country_market` arrays from all models for that deliverable.
  - Keep a clear, normalized **title** (no multipliers) and a normalized **category** string.
  - **Validate quantity**: Ensure technical_specifications × language_country_market ≈ quantity value.

---

## COMPLETENESS & COUNT CHECK
- Verify that every location/market/language found by any model appears (deduped) in the `language_country_market` array of the final deliverable.
- If overview sections imply the same base deliverable repeated across many locations, the final result should be **one base deliverable** with a populated `language_country_market` array whose length matches the unique values extracted.
- **Quantity validation**: Final expansion (technical_specifications × language_country_market) should approximately equal the `quantity` value.

---

## OUTPUT REQUIREMENTS
Return a JSON object with a single `"assets"` array containing the final set of **unique** BaseDeliverable objects with optimized multiplier structure. Each item must:
- Use the **normalized title** template (no multipliers in title).
- Use a **single normalized `category`** string.
- Include **only 2 multiplier arrays**: `technical_specifications` and `language_country_market`.
- Have `quantity` as **string** that validates the multiplication: technical_specifications × language_country_market ≈ quantity.
- Differ from all others by at least one **significant** data point (see above).

---

## EXAMPLES (generic)

**Example — per‑location titles collapse into one asset**
Model A:
{{
  "title": "Channel - Placement (Initiative)",
  "category": "Channel - Placement",
  "media": "IMAGE",
  "technical_specifications": ["As per supplied file"],
  "reference_material": "<link if present>",
  "brand_identifier": "Client Brand",
  "language_country_market": ["EN-Location-A", "EN-Location-B"],
  "quantity": "2"
}}
Model B:
{{
  "title": "1234 - Location A",
  "category": "Placement",
  "media": "IMAGE",
  "technical_specifications": ["As per supplied file"],
  "language_country_market": ["EN-Location-A"],
  "quantity": "1"
}}
**Result (merged)**:
{{
  "title": "Channel - Placement (Initiative)",
  "category": "Channel - Placement",
  "media": "IMAGE",
  "technical_specifications": ["As per supplied file"],
  "reference_material": "<link if present>",
  "brand_identifier": "Client Brand",
  "language_country_market": ["EN-Location-A", "EN-Location-B"],
  "quantity": "2"
}}

**Example — keep separate when file formats differ**
- Asset 1: `"asset_type":"JPG"`
- Asset 2: `"asset_type":"PNG"`
→ Significant difference → keep both; assign each the appropriate subset of multipliers.

---

## MODELS' ANALYSIS RESULTS

{models_results}

**TASK**: Consolidate these results into a single, comprehensive array of base deliverables following the strategy above.