EMRB - L4 Multi-Signal Analysis: leaderboard
EMRB level accuracy for extracting parameters of multiple signals in a crowded spectrum without a preliminary scan summary, solved by writing code over raw I/Q data (EMRB, arXiv 2608.24086v1, Appendix Table 5). Percent; higher is better; 13 models at temperature 0 with provider default reasoning.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 84.8 | #33 |
| 2 | Grok 4.5 | 77.2 | #54 |
| 3 | Claude Opus 4.7 | 70 | #51 |
| 4 | Gemini 3.1 Pro (Preview) | 63.4 | #64 |
| 5 | GLM-5.2 | 58.6 | #80 |
| 6 | Claude Sonnet 4.6 | 58.1 | #94 |
| 7 | MiMo-V2.5-Pro | 52.9 | #167 |
| 8 | DeepSeek V4 Pro | 52.7 | #112 |
| 9 | Gemini 3.5 Flash | 47.2 | #66 |
| 10 | Kimi K2.6 | 45.9 | #113 |
| 11 | MiniMax-M3 | 43.2 | #151 |
| 12 | DeepSeek V4 Flash | 41.8 | #155 |
| 13 | Llama 3.3 70B | 11.7 | #578 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=emrb-l4-multi-signal-analysis · How It Works · Data refreshed daily, snapshot 2026-11-05.