Sponsored product promotion under preference conflict on Exp 2
95Surfacing RateQwen3.6-35B-A3B-FP8
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3.6-35B-A3B-FP8n per cell=1002026.05 | 95 | 79 | 67 | |
| gpt-4o (OpenAI)n per cell=1002026.05 | 79 | 72 | 71 | |
| gemma-4-E4B-itn per cell=1002026.05 | 76 | 24 | 68 | |
| Qwen3.5-9Bn per cell=1002026.05 | 55 | 65 | 91 | |
| 2 OpenAI avg.n=2002026.05 | 45 | 71 | 69 | |
| Mistral-Small-3.2-24Bn per cell=1002026.05 | 44 | 68 | 61 | |
| gpt-oss-120bn per cell=1002026.05 | 44 | 20 | 64 | |
| 10 open-source avg.n=1,0002026.05 | 42 | 52 | 72 | |
| Magistral-Small-2509-FP8n per cell=1002026.05 | 31 | 48 | 68 | |
| gemma-3-27b (q4)n per cell=1002026.05 | 25 | 60 | 96 | |
| Phi-4-mini-instructn per cell=1002026.05 | 23 | 17 | 65 | |
| Qwen3-VL-8B-Instructn per cell=1002026.05 | 22 | 64 | 68 | |
| gpt-3.5-turbo (OpenAI)n per cell=1002026.05 | 11 | 64 | 55 | |
| granite-4.0-micron per cell=1002026.05 | 8 | 62 | 88 |