Arena's reward recipe lifts post-trained FLUX.2-dev by 69 Elo on live T2I leaderboard
arena · x · 2026-10-02
Arena details a post-training recipe for text-to-image models combining human preference rewards with rubric-based rewards: a Bradley-Terry reward model trained on 5.6M pairwise votes, a VLM-evaluated faithfulness reward from auto-generated checklists, constraint rewards covering explicit and implicit intent, and anti-reward-hacking rubrics targeting failures like garbled text and photorealism drift.
Results: post-trained FLUX.2-dev gains 69 Elo on Arena's live T2I leaderboard (1202); post-trained Ideogram 4 gains 20 Elo to 1224, surpassing all publicly listed open models. Offline ablations with Gemini 3.5 Flash as judge show complementary rewards push win rate to 64.2%, and weight-space ensembling of policies with/without the anti-reward-hacking objective raises it to 66.0%.
Related event: Arena Combines Preference and Rubric Rewards to Post-Train FLUX.2-dev(3 posts)→
More from Multimodal
- AI video brings Mona Lisa, Van Gogh and Dalí together across centuries — sebpaquet · 2026-11-19
- Fully Seedance 2.5-generated short shows AI video getting comedically coherent — anthara_ai · 2026-11-19
- Kling Image 4.0 debuts at #7 on Artificial Analysis, only model with native 8K output — anthara_ai · 2026-11-19
- Creator explains his slopcore video workflow built on a Claude plugin — AIandDesign · 2026-11-19
- Blender Bench: frontier models drive real Blender via MCP, humans blind-vote on 3D scenes — Gruku · 2026-11-19
- Core Memory profiles Ari Kuschnir, the creator behind viral AI parody videos — kyliebytes · 2026-11-19