{"id":"arxiv:2309.15807","title":"Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack","date":"2023-09-27","section":"Self-supervised learning and vision","quality":0.187565,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/","source_url":"https://arxiv.org/abs/2309.15807","text_available":true,"authors":["Xiaoliang Dai","Ji Hou","Chih-Yao Ma","Sam Tsai"],"outgoing":[],"incoming":[],"text_parts":[{"start":0,"end":15636,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html"},{"start":15636,"end":27168,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html"}],"contents":[{"title":"Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"Abstract","offset":81,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"1 Introduction","offset":1452,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"2 Related Work","offset":6463,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"Text-to-Image Models.","offset":6482,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"Fine-Tuning Text-to-Image Models.","offset":7378,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"Fine-Tuning Language Models.","offset":8097,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"3 Approach","offset":8713,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"3.1 Latent Diffusion Architecture","offset":10206,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"3.2 Pre-training","offset":11647,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"3.3 High-Quality Alignment Data","offset":12199,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-1.html#text"},{"title":"3.4 Quality-Tuning","offset":16975,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"4 Experiments","offset":17865,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"4.1 Evaluation Setting","offset":18556,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"Prompts.","offset":18584,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"Metrics.","offset":19318,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"4.2 Results","offset":20780,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"Effectiveness of Quality-Tuning.","offset":20797,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"Visual Appeal in the Context of SoTA.","offset":22504,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"Quality-Tuning Other Architectures.","offset":22836,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"4.3 Ablation Studies","offset":23494,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"5 Limitation","offset":23950,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"6 Conclusion","offset":25222,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"},{"title":"7 Acknowledgement","offset":26086,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/part-2.html#text"}],"text_code_points":27168,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f1b51d5d6f9eb66c9b6424518136746ec684fa58752ed4e2ec2a7031df9fe2ad/source.md"}