Semantic Consistency on UniGenBench In-domain v1
73.95Overall ScoreFLUX.1-dev w/ UnifiedReward-Flex
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FLUX.1-dev w/ UnifiedReward-FlexReward Model=UnifiedReward-Flex, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 73.95 | 90.3 | 89.87 | 79.38 | 73.38 | 78.55 | 69.46 | 63.1 | 46.59 | 79.66 | 59.2 | |
| FLUX.1-dev w/ UnifiedReward-ThinkReward Model=UnifiedReward-Think, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 68.89 | 88 | 91.77 | 77.99 | 69.2 | 75.13 | 61.47 | 61.63 | 41.36 | 77.24 | 45.11 | |
| FLUX.1-dev w/ UnifiedRewardReward Model=UnifiedReward, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 60.87 | 83.5 | 87.97 | 66.13 | 63.88 | 68.65 | 46.52 | 58.69 | 24.32 | 71.08 | 37.93 | |
| FLUX.1-devEval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 59.39 | 85.1 | 85.92 | 65.28 | 61.41 | 64.97 | 43.56 | 60.16 | 24.77 | 70.52 | 32.18 | |
| FLUX.1-dev w/ PickScoreReward Model=PickScore, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 58.63 | 79.7 | 87.03 | 64.42 | 61.12 | 67.64 | 47.42 | 58.02 | 27.5 | 67.54 | 25.86 | |
| FLUX.1-dev w/ HPSv3Reward Model=HPSv3, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 57.98 | 79.4 | 90.03 | 66.24 | 57.89 | 63.58 | 39.82 | 58.82 | 24.09 | 67.16 | 32.76 | |
| FLUX.1-dev w/ HPSv2Reward Model=HPSv2, Eval Model=UniGenBench-EvalModel-qwen3vl-32b-v12026.02 | 57.77 | 77.9 | 87.03 | 65.92 | 57.41 | 65.86 | 44.46 | 55.75 | 29.09 | 64.93 | 29.31 |