Automatically extracting effective queries is challenging in information retrieval, especially in toxic content exploration, as such content is likely to be disguised. With the recent achievements in generative Large Language Model (LLM), we are able to leverage the capabilities of LLMs to extract effective queries for similar content exploration directly. This study proposes QExplorer, an approach of large language model based Query Extraction for toxic content Exploration. The QExplorer approach involves a 2-stage training process: instruction Supervised FineTuning (SFT) and preference alignment using Direct Preference Optimization (DPO), as well as the datasets construction with feedback of search system. To verify the effectiveness of QExplorer, a series of offline and online experiments are conducted on our real-world system. The offline empirical results demonstrate that the performance of our automatic query extraction outperforms that of several LLMs and humans. The online deployment shows a significant increase in the detection of toxic items.

Despite notable advancements in Retrieval-Augmented Generation (RAG) systems that expand large language model (LLM) capabilities through external retrieval, these systems often struggle to meet the complex and diverse needs of real-world industrial applications. The reliance on retrieval alone proves insufficient for extracting deep, domain-specific knowledge performing in logical reasoning from specialized corpora. To address this, we introduce sPecIalized KnowledgE and Rationale Augmentation Generation (PIKE-RAG), focusing on extracting, understanding, and applying specialized knowledge, while constructing coherent rationale to incrementally steer LLMs toward accurate responses. Recognizing the diverse challenges of industrial tasks, we introduce a new paradigm that classifies tasks based on their complexity in knowledge extraction and application, allowing for a systematic evaluation of RAG systems' problem-solving capabilities. This strategic approach offers a roadmap for the phased development and enhancement of RAG systems, tailored to meet the evolving demands of industrial applications. Furthermore, we propose knowledge atomizing and knowledge-aware task decomposition to effectively extract multifaceted knowledge from the data chunks and iteratively construct the rationale based on original query and the accumulated knowledge, respectively, showcasing exceptional performance across various benchmarks.

While I am only guessing that performance may degrade as a function of dataset scale, it is not hard to imagine advances in GANs which could make that degradation smaller, hence make the proposed method more useful. Further, even in an adversarial setting, it may be possible to guess what kind of inputs are relevant, or extend the method to few-shot or some hybrid approach. I am positively surprised that features of the student have comparable transferability to the teacher, I was concerned that some sort of overfitting to a teacher's decision boundary was possible, but this does not seem to be the case. While I agree with the authors that, in most cases, those releasing research models will not go out of their way to vaccinate them against zero-shot distillation, the proposed method could be used to (somewhat) copy and repurpose information stored in hardware model. Take for example Tesla's autopilot which uses several neural networks in it and is trained on tens of billions of images which are not available to the world.

All reviewers appreciated the work and recommend acceptance. The authors are encouraged to address the reviewers comments and include the author response in the camera-ready version of the manuscript.

OpenAI is showing no signs of slowing down its recent pace of updates. On Wednesday, the company announced the expanded availability of ChatGPT Search. After rolling out the tool first to paid subscribers last fall, and then making it available to all logged-in free users at the end of 2024, now anyone can use ChatGPT Search with no account or sign-in necessary. "Like the logged-in experience, ChatGPT can search the web and get you fast, timely answers with links to relevant web sources directly in ChatGPT," OpenAI said. In most cases, ChatGPT will automatically search the web to source the most up-to-date information related to your question.

The release of GPT-3, and later ChatGPT, catapulted large language models from the proceedings of computer science conferences to newspaper headlines across the globe, fueling their rise to one of today's most hyped technologies. The public's awe about GPT-3's knowledge and fluency was quickly blemished by concerns regarding its potential to radicalize, instigate, and misinform, for example, by stating that Bill Gates aimed to "kill billions of people with vaccines" or that Hillary Clinton was a "high-level satanic priestess."4 These shortcomings, in turn, have sparked a surge in research on AI alignment,7 a field aiming to "steer AI systems toward a person's or group's intended goals, preferences, and ethical principles" (definition by Wikipedia). A well-aligned AI system will "understand" what is "good" and what is "bad" and will do only the "good" while avoiding the "bad."a The resulting techniques, including instruction fine-tuning, reinforcement learning from human feedback, and so forth, have contributed in major ways to improving the output quality of large language models.

LinkedIn is testing a new job-hunting tool that uses a custom large language model to comb through huge quantities of data to help people find prospective roles. The company believes that artificial intelligence will help users unearth new roles they might have missed in the typical search process. "The reality is, you don't find your dream job by checking a set of keywords," the company's CEO, Ryan Roslansky, told WIRED in a statement. The new tool, he says, "can help you find relevant jobs you never even knew to search for." The move comes as AI continues to change how people use the web.

INNOVATIVE STEP FORWARD: Syntilay, a startup with a sparkle in its eye and artificial intelligence on its mind, has just unveiled what it claims to be the world's first entirely AI-designed and 3D-printed shoe. SETTING THE RECORD STRAIGHT: President Donald Trump's artificial intelligence czar, David Sacks, is pointing to evidence that China's DeepSeek AI startup spent a lot more money developing its models than has been reported. ROBOT'S GOT MOVES: Deep Robotics, a Chinese robotics firm, recently unveiled its latest innovation in quadruped robotics, the Lynx. SPUTNIK MOMENT: If you care about national security, artificial intelligence (AI) or the index funds in your retirement account, you have likely heard of DeepSeek. Chinese AI model DeepSeek's release late January caused a 969 billion stock market selloff and prompted responses from AI leaders like President Donald Trump, NVIDIA, venture capitalist Marc Andreessen and OpenAI CEO Sam Altman.

Microsoft's annual Build developer conference will take place in Seattle and run from May 19 to May 22, the company announced on X. There are no details on what will be announced, but you can be fairly sure it'll match or beat Build 2024 in terms of AI-related products and services. Artificial intelligence played a large part in last year's conference, featuring in areas ranging from Windows search to Copilot to Microsoft Paint. In fact, a day before Build 2024, Microsoft unveiled its new Surface Pro PC powered by Qualcomm's Snapdragon X Elite and X Plus chips, capable of hitting 45TOPS of neural processing power. Come join us at #MSBuild, May 19 – 22, 2025.

Australia has banned DeepSeek AI services from all government systems and devices, becoming one of the first countries to take direct action against a Chinese artificial intelligence startup that shook Silicon Valley and global markets this year. Home Affairs Minister Tony Burke said in a statement Tuesday that all DeepSeek products, applications and services would be removed from government systems on national security grounds effective immediately. A threat assessment by the country's intelligence agencies found the technology posed an unacceptable risk, he said. Founded in Hangzhou only 20 months ago, DeepSeek's technology made waves in January with a new mobile app featuring its reasoning AI chatbot -- which articulates its approximation of thought process and research before delivering a response -- that seemed to suggest top-tier AI could be developed without huge investments in hardware. Its appeal took it to the top of worldwide download charts.