RPCBench - Error Group - Unverifiable Premise - Evidence Distortion: leaderboard
Metric: Evidence Distortion Rate (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 3.63 |
| 2 | Qwen 3.5 397B A17B | 8.53 |
| 3 | Gemini 3.1 Pro (Preview) | 8.92 |
| 4 | Qwen 3.5 122B A10B | 9.8 |
| 5 | Qwen 3.5 35B A3B | 10.29 |
| 6 | Qwen 3.5 Plus | 10.39 |
| 7 | DeepSeek V4 Flash | 11.67 |
| 8 | DeepSeek V4 Pro | 12.06 |
| 9 | Llama 3.1 8B Instruct | 12.94 |
| 10 | Claude Sonnet 4.6 | 13.53 |
| 11 | Llama 3.1 70B Instruct | 17.45 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-unverifiable-premise-evidence-distortion · How It Works · Data refreshed daily, snapshot 2026-09-24.