[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"aihot-art-87242":3},{"itemId":4,"vertical":5,"category":6,"source":7,"score":8,"title":9,"summary":10,"analysis":11,"url":12,"coverUrl":13,"direction":13,"marketSignal":13,"publishedAt":14},"87242","ai","论文研究","AK",56,"预训练多模态大语言模型作为文生图的零样本奖励模型","研究提出预训练多模态大语言模型可作为文生图的零样本奖励模型。","预训练多模态大语言模型可作为文生图的零样本奖励模型。\n· 研究发现，预训练 MLLM 无需额外训练即可作为奖励模型，用于文本到图像生成。\n· 零样本能力意味着可直接应用于未见过的任务，无需微调。\n· 该方法有望提升文生图模型的对齐质量和生成效果。\n· 为利用现有大模型改进生成任务提供了新思路。\n影响\u002F看点：这一发现简化了文生图系统的优化流程，可能推动更高质量、更可控的图像生成应用。","https:\u002F\u002Fx.com\u002F_akhaliq\u002Fstatus\u002F2077420923410358343",null,"2026-07-15 23:51:59"]