List your company

Scene Generation on World Craft (test)

94CFRTwo-Stage + Critic (Ours)
57.667.0576.585.95Jan 14, 2026

Evaluation results

MethodLinks
Two-Stage + Critic (Ours)Data=Correction Data2026.0194883.03997.133.6428.076.8
Training Variants (w/o Critic)Strategy=Stage 2 Only (Manager)2026.0189784.11916.844.7622.164.34
Two-Stage + CriticData=Standard Data2026.0189813.67947.074.3926.836.04
Training Variants (w/o Critic)Backbone=Qwen3-(8+32)B2026.0188793.93927.14.7626.366.17
Training Variants (w/o Critic)Strategy=End-to-End Fine-tuning2026.0186785.36876.124.8124.835.33
Direct Gen. + CriticFramework=Direct Generation + Critic, Backbone=Gemini-3-Pro2026.0184674.99905.35.9223.514.98
Enricher + Manager + CriticFramework=Enricher + Manager + Critic, Backbone=Gemini-3-Pro2026.0183683.77925.925.925.085.71
Training Variants (w/o Critic)Backbone=Qwen3-(8+8)B2026.0183734.64906.355.4225.315.64
Enricher + Manager + CriticFramework=Enricher + Manager + Critic, Backbone=Qwen3-235B2026.0181714.72845.286.1324.725.66
Direct Gen. (Few-shot)Framework=Direct Generation, Inference Mode=Few-shot, Backbone=Gemini-3-Pro2026.0180595.56865.026.8723.294.67
Direct Gen. + CriticFramework=Direct Generation + Critic, Backbone=Qwen3-235B2026.0178676.88834.376.1723.735.03
Direct Gen. (Few-shot)Framework=Direct Generation, Inference Mode=Few-shot, Backbone=Qwen3-235B2026.0173607.18804.097.2223.344.74
Direct Gen. + CriticFramework=Direct Generation + Critic, Backbone=Qwen3-32B2026.0166498.68783.439.3120.243.88
Enricher + Manager + CriticFramework=Enricher + Manager + Critic, Backbone=Qwen3-(8+32)B2026.0166527.47774.469.3321.454.03
Training Variants (w/o Critic)Strategy=Stage 1 Only (Enricher)2026.0163468.16763.979.8825.225.22
Direct Gen. (Few-shot)Framework=Direct Generation, Inference Mode=Few-shot, Backbone=Qwen3-32B2026.0159449.27763.2710.5520.223.73