A | 8月26日消息，SuperCLUE Image放出8月份中文文生图测评榜单。 这次一共测了14款国内外主流模型，还新增复杂信息布局这个考核维度，专门看模型做信息图、分镜、UI界面的能力。 A former Ohio State University student-athlete is speaking out against Rep. Jim Jordan's bid for House speaker, saying Jordan "turned a blind eye" to allegations of abuse against a team doctor during his time as assistant coach for the university's wrestling team.Rocky Ratliff is now an attorney who also represents several other former OSU wrestlers in an ongoing lawsuit against the university."I think the wrestlers that I represent, not one of us, would back him for such a leadership position," Ratliff told ABC News Live Prime anchor Linsey Davis.He continued, "He's abandoned us for his own selfish reasons when he could have helped us. He's chosen not to. So that is not the good makings of any type of leadership or any type of leader that he would have put up with at Ohio State. It's just not. None of us wrestlers believe he should get that position."A spokesperson for Jordan, R-Ohio, told ABC News, "Chairman Jordan never saw or heard of any abuse, and if he had, he would have dealt with it."The allegations of abuse were against Dr. Richard Strauss, an Ohio State team doctor and sports medicine researcher. Strauss is accused of sexually abusing at least 177 men over an 18-year period from 1979 to 1997 -- nearly his entire time at Ohio State, according to an independent report released in 2019.Investigators determined that university officials ignored nearly two decades of accusations of sexual abuse against Strauss, who killed himself in 2006 at age 67, seven years after retiring from the university.The accusations of abuse involved athletes from at least 16 sports including wrestling, hockey and swimming, and included Strauss' work at the student health center and an off-campus clinic that he founded late in his tenure, according to the report.The university removed Strauss as a school physician in 1996 after a flurry of student complaints and reported his actions to the State Medical Board of Ohio. However, the school allowed Strauss to retain his tenured faculty position while he operated an off-campus clinic, where the report says he continued to abuse students.Ohio State University has since admitted that it failed to protect students from Strauss, paying out $60 million in settlements to some 296 victims.Jordan, who was an assistant coach on the team from 1986 to 1994, came under fire in 2018 when several former OSU wrestlers took their allegations against Strauss to the media and claimed Jordan was aware of Strauss' inappropriate behavior and failed to report it. The university then announced it was opening an investigation into the allegations against Strauss.Jordan also denied knowing about the abuse when the allegations first came out in 2018.Jordan is now facing renewed scrutiny amid the ongoing battle for House speaker. After again failing to receive enough GOP support, the congressman and chair of the House Judiciary Committee was defeated a second time after a vote on Wednesday afternoon."We believe very strongly, especially all the wrestlers that were there at the time, that Jim Jordan knew what was going on," Ratliff said on Wednesday.Ratliff continued, "Jordan should come forward and tell the truth about what happened. At least meet with the guys. He's failed to do that -- to hear our side, he's failed. You know, even if you believe what Jim Jordan says, he has never once reached out to any wrestler to say, 'Hey, I missed it. I'm sorry. How are you feeling?'"ABC News' Eric Ortega, Imtiyaz Delawala, Lindsey Griswold, and Andrea Amiel contributed to this report.。 这次榜单最大看点，商汤SenseNova U1 Pro拿到93.81分，以0.58分的微弱优势超过OpenAI GPT Image 2，直接拿下全球总榜第一。 字节跳动Doubao Seedream 5.0 pro、阿里Qwen Image 3.0 Pro紧随其后。总榜前五里面，国内模型占了3个。
B | 从得分情况能看出来，现在不少模型做画面还原、创意生成已经很成熟，这两项平均分接近90分。 真正拉开差距的是图文一致性，还有刚新增的复杂信息布局，两项平均分只有72分出头。
C | 特别是复杂信息布局，最高分和最低分差74.2分，是分化最厉害的一项。
D | 简单说，生成好看图片的大模型很多，但能在一张图里搞定大量文字、表格、流程图的模型很少。
E | 榜单也能看出明显两极分化。
F | 排名前2的商汤、OpenAI模型，六大测评维度全部站在第一梯队，属于全能选手。 但很多中尾部模型偏科很严重，某一项能力很强，别的维度直接拉胯，只靠单个亮点已经冲不上高排名。 国产模型不是所有项目都占优。汉字生成、现实复现这两块国内表现更好，可在图文一致性上，和海外模型还有16.56分的不小差距。 总的来说，国产头部文生图已经站上全球第一的位置，但不是整体全部领先。
G | 能不能处理复杂图文信息，已经变成判断文生图模型好不好用的核心标准。
Current article:http://v31s.xuebenwagangqingre.buzz/r3i/20260826/219776.html
Published on:02:53:48
