RPCBench - Evidence Faithfulness: leaderboard
RPCBench tests proactive critique of flawed recommendation premises, locating the premise problem and suggesting a suitable strategy while remaining faithful to the provided evidence. This series reports RPCBench - Evidence Faithfulness from 2609.00918v1, S3.T2: Faithfulness Index (%). Higher is better.
Metric: Faithfulness Index (%). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 87.9 | #34 |
| 2 | Qwen 3.5 122B A10B | 87.6 | #186 |
| 3 | Qwen 3.5 Plus | 87.4 | #136 |
| 4 | Qwen 3.5 397B A17B | 87.3 | #155 |
| 5 | Qwen 3.5 35B A3B | 86.5 | #270 |
| 6 | Gemini 3.1 Pro (Preview) | 83.6 | #66 |
| 7 | DeepSeek V4 Pro | 79.7 | #113 |
| 8 | DeepSeek V4 Flash | 78.5 | #156 |
| 9 | Claude Sonnet 4.6 | 74.7 | #97 |
| 10 | Llama 3.1 70B Instruct | 40.9 | #570 |
| 11 | Llama 3.1 8B Instruct | 34.7 | #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-evidence-faithfulness · How It Works · Data refreshed daily, snapshot 2026-11-07.