# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-163314
**Date:** Apr 02, 2026
**Category:** reasoning
**Question ID:** REASON-013

---

## Question

Sleeping Beauty is told: 'We'll flip a coin. If heads, we wake you once (Monday). If tails, we wake you twice (Monday and Tuesday) with memory erased between wakings. Each time you wake, we'll ask: what is your credence that the coin landed heads?' (1) Is the answer 1/2 (halfer position) or 1/3 (thirder position)? (2) Argue for both positions rigorously. (3) What if we change the experiment to 1 waking for heads and 1,000,000 wakings for tails?

---

## Winner

**Grok 4.20** (openrouter)
- Winner Score: 8.90
- Matrix Average: 8.07
- Total Judgments: 88

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | Grok 4.20 | openrouter | 8.90 | 9 |
| 2 | Claude Sonnet 4.6 | openrouter | 8.90 | 8 |
| 3 | GPT-5.4 | openrouter | 8.86 | 9 |
| 4 | MiniMax M2.5 | openrouter | 8.81 | 9 |
| 5 | Claude Opus 4.6 | openrouter | 8.79 | 9 |
| 6 | MiMo-V2-Flash | Xiaomi | 8.57 | 9 |
| 7 | GPT-OSS-120B | OpenAI | 8.53 | 8 |
| 8 | Gemini 2.5 Flash | openrouter | 7.32 | 9 |
| 9 | DeepSeek V4 | openrouter | 6.54 | 9 |
| 10 | Gemini 3.1 Pro | openrouter | 5.50 | 9 |

---

## 10×10 Judgment Matrix

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

| Judge ↓ / Resp → | DeepSeek V4 | Gemini 3.1 Pro | Claude Opus | GPT-5.4 | Grok 4.20 | Claude Sonnet | MiMo-V2-Flash | GPT-OSS-120B | Gemini 2.5 | MiniMax M2.5 |
|---|---|---|---|---|---|---|---|---|---|---|
| DeepSeek V4 | — | 8.4 | 9.7 | 9.4 | 9.4 | 9.4 | 9.3 | 9.7 | 8.8 | 9.3 |
| Gemini 3.1 Pro | 5.8 | — | 8.7 | 8.1 | 9.8 | 9.3 | 9.4 | · | 6.5 | 9.4 |
| Claude Opus | 5.7 | 4.5 | — | 8.4 | 8.8 | 8.2 | 7.8 | 7.2 | 6.7 | 8.6 |
| GPT-5.4 | 4.2 | 2.6 | 7.7 | — | 7.8 | · | 8.2 | 6.6 | 4.5 | 8.0 |
| Grok 4.20 | 6.8 | 7.9 | 8.7 | 8.8 | — | 8.7 | 8.7 | 8.7 | 8.3 | 8.7 |
| Claude Sonnet | 5.7 | 4.5 | 9.0 | 9.0 | 9.2 | — | 8.6 | 8.7 | 8.0 | 8.6 |
| MiMo-V2-Flash | 8.4 | 4.2 | 8.8 | 9.0 | 9.0 | 9.4 | — | 9.4 | 8.8 | 9.0 |
| GPT-OSS-120B | 5.8 | 3.8 | 8.7 | 8.7 | 8.3 | 8.4 | 7.4 | — | 6.7 | 8.3 |
| Gemini 2.5 | 8.8 | 8.4 | 9.1 | 8.8 | 9.4 | 8.8 | 9.4 | 8.8 | — | 9.4 |
| MiniMax M2.5 | 7.6 | 5.1 | 8.8 | 9.4 | 8.4 | 9.0 | 8.4 | 9.3 | 7.7 | — |

---

## 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: REASON-013. 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-20260402-163314/results
Full dataset: https://app.themultivac.com/dashboard/export