Large Language Models (LLMs) as judges and LLM-based data synthesis have emerged as two fundamental LLM-driven data annotation methods in model development. While their combination significantly enhances the efficiency of model training and evaluation, little attention has been given to the potential contamination brought by this new model development paradigm. In this work, we expose preference leakage, a contamination problem in LLM-as-a-judge caused by the relatedness between the synthetic data generators and LLM-based evaluators. To study this issue, we first define three common relatednesses between data generator LLM and judge LLM: being the same model, having an inheritance relationship, and belonging to the same model family. Through extensive experiments, we empirically confirm the bias of judges towards their related student models caused by preference leakage across multiple LLM baselines and benchmarks. Further analysis suggests that preference leakage is a pervasive issue that is harder to detect compared to previously identified biases in LLM-as-a-judge scenarios. All of these findings imply that preference leakage is a widespread and challenging problem in the area of LLM-as-a-judge. We release all codes and data at: https://github.com/David-Li0406/Preference-Leakage.

Mae Sot, Thailand – Once again, the attack came from the sky. The Kachin resistance fighters barely heard the sound of the propellers as the Myanmar military's two drones released their payload above their heads in northern Kachin State in late April. "I fell down to the ground when the bombs dropped," Aung Nge, a fighter with the Kachin People's Defense Force (PDF), told Al Jazeera from an undisclosed location. I was awake the whole time." The drone attack seriously injured three men who were holed up close to the front line in Kachin State where battles with the armed forces have been escalating since October last year. In critical condition, field medics sent the men to a hidden hospital deep in the jungle where they could be treated by professional doctors. Within a day of receiving treatment, however, one of the soldiers started to show symptoms the doctors could not understand and his condition began to deteriorate rapidly. Another man from the Kachin Independence Army (KIA), who had been injured in a separate drone strike days after the first attack and appeared to be on the mend with no signs of infection, also took a turn for the worse and died in his sleep. Aung Nge, meanwhile, was about to endure ghastly infections that would spread across his entire body. Doctors told Al Jazeera that the men experienced rapid onset necrosis, an effect not normally seen in a blast wound. Necrosis causes the deterioration of most or all of the cells in an organ or tissue due to disease or the failure of the blood supply. While necrosis can be caused by sepsis, which appears rapidly and is usually accompanied by a fever, doctors said they could find no physiological reason for the rapid deterioration in their patients. Toxic substances can also trigger such reactions, they said. "In close examination of the wounds, they are rapidly necrotising, easily decomposed and not associated with metallic foreign bodies," Dr Soe Min, the veteran trauma doctor who treated the suspicious cases, told Al Jazeera. He has been treating combat-related cases since January 2022 and has seen and treated hundreds of blast injuries. These cases were different, he said. "After two days, all the wounds became blackish in colour with foul-smelling discharge.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. BANGKOK (AP) -- Myanmar's main pro-democracy resistance group said Thursday its armed wing launched drone attacks on the airport and a military headquarters in the capital, Naypyitaw, but the country's ruling military said it destroyed or seized more than a dozen drones used in the attacks. The opposition National Unity Government's "Defense Ministry" said in a statement that special units of the People's Defense Force used drones to attack the targets simultaneously. The group, known by the acronym NUG, calls itself the country's legitimate government, while the People's Defense Force is made up of many local resistance groups with a good deal of independence.

Myanmar's main pro-democracy opposition group says its armed wing has launched drone attacks on an airport and a military headquarters in the capital, Naypyidaw, but the military government says it has destroyed or seized more than a dozen drones used in the attacks. Military-run Myawaddy TV reported on Thursday that 13 fixed-wing drones were shot down over the capital of military-ruled Myanmar in a foiled attack with no casualties or damage to property. It said the foiled attack by "terrorists" sought to destroy important locations in Naypyidaw. Myawaddy did not mention what the targets were but broadcast an image showing nine small drones, several of which were damaged. Of the 13 drones, four carried explosives, the report said.

Rapid advances in artificial intelligence are boosting online disinformation and enabling governments to increase censorship and surveillance in a growing threat to human rights, a U.S. nonprofit said in a report published Wednesday. Global internet freedom declined for the 13th consecutive year, with China, Myanmar and Iran having the worst conditions of the 70 countries surveyed by the Freedom on the Net report, which highlighted the risks posed by easy access to generative AI technology. AI allows governments to "enhance and refine online censorship" and amplify digital repression, making surveillance, and the creation and spread of disinformation faster, cheaper, and more effective, said the annual report by Freedom House.

Global internet freedom declined for a thirteenth consecutive year in 2023, partially as a result of AI being used to sow disinformation and enhance content censorship, according to a new report from U.S.-based nonprofit Freedom House. The 2023 Freedom on the Net report, published on Oct. 4, assesses the state of internet freedom in 70 countries through a comprehensive methodology examining obstacles to access, limits on content, and violations of user rights. The report found that many countries--including Myanmar, the Philippines, Costa Rica--have drastically restricted online freedoms this year. China has the lowest levels of internet freedom for the ninth consecutive year, the report said. Freedom House, established in 1941, publishes Freedom in the World and Freedom on the Net annually.

"We needed weapons, and we needed them fast," 3D says, sitting beneath the stalactites in a dimly lit cave, somewhere deep in the jungle in eastern Myanmar. The space reverberates with the hum of 3D printers--the devices that gave 3D his nom de guerre. He spoke on condition that WIRED would not reveal his real name or show his face. "My parents would kill me if they would know what I'm up to," he says. Not only does 3D face the risk of arrest, torture, or execution for his part in the revolution, the military would not hesitate to arrest his parents if they were to discover 3D's identity.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. A drone attack on a police headquarters in a major border town in eastern Burma has killed at least five officials including a senior army officer and a district administrator, members of two emergency rescue teams and media reports said Monday. The attack, carried out Sunday evening in two stages, is believed to be the deadliest aerial bombing targeting high-ranking security and administrative officials since armed resistance was launched more than two years ago against the military that seized power in February 2021 from the elected government of Aung San Suu Kyi. The takeover was met with peaceful nationwide protests, but after security forces cracked down with lethal force, many local armed resistance groups were formed and loosely organized into what is called the People's Defense Force, or PDF. It's the armed wing of Burma's shadow National Unity Government, which views itself as a country's legitimate administrative body.

This paper presents a state-of-the-art model for transcribing speech in any language into the International Phonetic Alphabet (IPA). Transcription of spoken languages into IPA is an essential yet time-consuming process in language documentation, and even partially automating this process has the potential to drastically speed up the documentation of endangered languages. Like the previous best speech-to-IPA model (Wav2Vec2Phoneme), our model is based on wav2vec 2.0 and is fine-tuned to predict IPA from audio input. We use training data from seven languages from CommonVoice 11.0, transcribed into IPA semi-automatically. Although this training dataset is much smaller than Wav2Vec2Phoneme's, its higher quality lets our model achieve comparable or better results. Furthermore, we show that the quality of our universal speech-to-IPA models is close to that of human annotators.

The growing potential of generative AI is clouded by its possible harms, prompting some calls for regulation. ChatGPT and other generative AI have taken centerstage for innovation with companies racing to introduce their own respective twists on the technology. Questions about the ethics of AI have likewise escalated with ways the technology could spread misinformation, support hacking attempts, or raise doubts about the ownership and validity of digital content. The issue of ethics and AI is not new, according to Cynthia Rudin, the Earl D. McLean, Jr. professor of computer science, electrical and computer engineering, statistical science, mathematics, and biostatistics & bioinformatics at Duke University She says AI recommender systems already have been pointed to for such ills as contributing to depression among teenagers, algorithms amplifying hate speech that spurred the 2017 Rohingya massacre in Myanmar, vaccine misinformation, and the spread of propaganda that contributed to insurrection in the United States on January 6, 2021. "If we haven't learned our lesson about ethics by now, it's not going to be when ChatGPT shows up," says Rudin.