CL-Bench Sales Prediction Gain
Agents · 2026-05-04
Mean cumulative gain over the stateless baseline across 12 sequential instances of the CL-Bench sales prediction task.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 4.6 | 5.8 |
| Gemini 3.1 Pro Preview | 5.0 |
| Opus 4.7 | 4.6 |
| GPT-5.4 | 4.0 |
| Gemini 3 Flash Preview | 3.2 |