RPCBench - Error Group - Boundary-Violating Premise - Evidence Faithfulness: leaderboard
RPCBench tests proactive premise critique in recommendation requests. This series measures evidence faithfulness on the Boundary-Violating Premise subset (436 questions), using the same eleven-model, three-judge experiment as the headline paper. Published EFI fractions from the authors' per-model CSV are scaled once to Faithfulness Index (%). Higher is better; domains and nested premise-error groups are related slices of the same experiment.
Metric: Faithfulness Index (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Plus | 99.54 | #136 |
| 2 | Qwen 3.5 397B A17B | 99.2 | #155 |
| 3 | GPT-5.5 | 98.39 | #34 |
| 4 | Qwen 3.5 122B A10B | 98.17 | #186 |
| 5 | Qwen 3.5 35B A3B | 95.53 | #270 |
| 6 | Gemini 3.1 Pro (Preview) | 92.43 | #66 |
| 7 | Claude Sonnet 4.6 | 91.17 | #97 |
| 8 | DeepSeek V4 Pro | 82.22 | #113 |
| 9 | DeepSeek V4 Flash | 81.31 | #156 |
| 10 | Llama 3.1 70B Instruct | 70.64 | #570 |
| 11 | Llama 3.1 8B Instruct | 62.61 | #1021 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-boundary-violating-premise-evidence-faithfulness · How It Works · Data refreshed daily, snapshot 2026-11-07.