# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-123136
**Date:** Feb 17, 2026
**Category:** code
**Question ID:** CODE-006

---

## Question

Write comprehensive unit tests for this function. Cover all edge cases, including boundary conditions, error cases, and typical usage.

```python
def merge_sorted_streams(*streams, max_items=None):
    """
    Merge multiple sorted iterables into a single sorted output.
    
    Args:
        *streams: Variable number of sorted iterables
        max_items: Optional limit on total items to yield
    
    Yields:
        Items from all streams in sorted order
    
    Raises:
        ValueError: If any stream is not sorted
    """
    import heapq
    
    heap = []
    iterators = [iter(s) for s in streams]
    
    # Initialize heap with first item from each stream
    for i, it in enumerate(iterators):
        try:
            item = next(it)
            heapq.heappush(heap, (item, i))
        except StopIteration:
            pass
    
    count = 0
    prev = None
    
    while heap and (max_items is None or count < max_items):
        item, stream_idx = heapq.heappop(heap)
        
        # Validate sorting
        if prev is not None and item < prev:
            raise ValueError(f"Stream {stream_idx} is not sorted")
        
        yield item
        prev = item
        count += 1
        
        # Get next item from same stream
        try:
            next_item = next(iterators[stream_idx])
            heapq.heappush(heap, (next_item, stream_idx))
        except StopIteration:
            pass
```

Use pytest. Include parametrized tests where appropriate.

---

## Winner

**GPT-5.4** (openrouter)
- Winner Score: 9.08
- Matrix Average: 7.20
- Total Judgments: 86

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | GPT-5.4 | openrouter | 9.08 | 9 |
| 2 | DeepSeek V4 | openrouter | 8.37 | 9 |
| 3 | Gemini 3 Flash Preview | Google | 8.33 | 8 |
| 4 | Grok 4.20 | openrouter | 7.78 | 9 |
| 5 | MiMo-V2-Flash | Xiaomi | 7.56 | 9 |
| 6 | GPT-OSS-120B | OpenAI | 7.23 | 9 |
| 7 | Claude Opus 4.6 | openrouter | 6.88 | 8 |
| 8 | Claude Sonnet 4.6 | openrouter | 6.71 | 7 |
| 9 | MiniMax M2.5 | openrouter | 6.61 | 9 |
| 10 | Gemini 3.1 Pro | openrouter | 3.40 | 9 |

---

## 10×10 Judgment Matrix

Rows = Judge, Columns = Respondent. Self-judgments excluded (—).

| Judge ↓ / Resp → | GPT-5.4 | Claude Opus | Gemini 3.1 Pro | Claude Sonnet | Grok 4.20 | DeepSeek V4 | GPT-OSS-120B | Gemini 3 | MiniMax M2.5 | MiMo-V2-Flash |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-5.4 | — | 3.8 | 0.7 | 2.6 | 5.3 | 7.3 | 2.9 | 7.1 | 2.6 | 4.5 |
| Claude Opus | 9.0 | — | 1.6 | 7.7 | 7.6 | 8.0 | 6.8 | 7.5 | 5.3 | 7.7 |
| Gemini 3.1 Pro | 9.6 | 6.4 | — | 4.8 | 5.5 | 9.3 | 6.0 | 9.4 | 5.3 | 7.8 |
| Claude Sonnet | 9.0 | 8.0 | 1.2 | — | 8.0 | 8.3 | 7.3 | 8.2 | 7.0 | 8.0 |
| Grok 4.20 | 8.6 | 6.8 | 3.6 | · | — | 6.6 | 8.3 | 8.6 | 7.7 | 7.5 |
| DeepSeek V4 | 9.6 | 8.6 | 8.2 | 8.6 | 9.3 | — | 9.1 | 9.3 | 9.6 | 8.8 |
| GPT-OSS-120B | 8.8 | 6.2 | 2.2 | · | 7.5 | 8.8 | — | 8.8 | 6.5 | 7.5 |
| Gemini 3 | 10.0 | · | 2.9 | 7.8 | 9.6 | 9.8 | 9.0 | — | 6.8 | 9.2 |
| MiniMax M2.5 | 8.8 | 7.5 | 1.2 | 7.5 | 8.2 | 8.6 | 7.3 | · | — | 7.2 |
| MiMo-V2-Flash | 8.3 | 7.8 | 9.0 | 8.0 | 8.8 | 8.6 | 8.6 | 7.8 | 8.6 | — |

---

## Methodology

- **10×10 Blind Peer Matrix:** All models answer the same question, then all models judge all responses.
- **5 Criteria:** Correctness, completeness, clarity, depth, usefulness (each scored 1–10).
- **Self-judgments excluded:** Models do not judge their own responses.
- **Weighted Score:** Composite of all 5 criteria.

---

## Citation

The Multivac (2026). Blind Peer Evaluation: CODE-006. app.themultivac.com

## License

Open data. Free to use, share, and build upon. Please cite The Multivac when using this data.

Download raw JSON: https://app.themultivac.com/api/evaluations/EVAL-20260402-123136/results
Full dataset: https://app.themultivac.com/dashboard/export