RPCBench - Composite Premise Critique: 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 - Composite Premise Critique from 2609.00918v1, S3.T2: Composite Score (0-100). Higher is better.
Metric: Composite Score (0-100). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Plus | 52.61 | #232 |
| 2 | Qwen 3.5 397B A17B | 52.59 | #245 |
| 3 | Qwen 3.5 122B A10B | 52.45 | #335 |
| 4 | GPT-5.5 | 51.8 | #48 |
| 5 | DeepSeek V4 Pro | 51.66 | #188 |
| 6 | Claude Sonnet 4.6 | 50.92 | #159 |
| 7 | Gemini 3.1 Pro (Preview) | 50.47 | #93 |
| 8 | Qwen 3.5 35B A3B | 48.51 | #492 |
| 9 | DeepSeek V4 Flash | 46.55 | #259 |
| 10 | Llama 3.1 70B Instruct | 13.59 | #1629 |
| 11 | Llama 3.1 8B Instruct | 10.19 | #4318 |
No result here: #4 Claude Opus 5.5, #5 GPT-6 Astra, #9 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-composite-premise-critique · How It Works · Data refreshed daily, snapshot 2026-11-08.