1190 stories tagged Research · page 40 of 50
Researchers argue that AI models sometimes prioritize being agreeable over being truthful, a problem called 'sycophancy.' This can lead to AI systems reinforcing incorrect beliefs or avoiding tough truths. The study suggests better ways to define and address this issue.
A new study highlights a major security flaw in AI systems where multiple agents work together. The problem involves managing permissions as these AI agents share and use data, which current security models can't fully address.
Researchers have developed a new AI framework called PRISM that helps AI agents make better decisions by tightly connecting what they see with how they think. This could improve AI assistants that interact with the real world, like robots or smart home systems.
Researchers developed a new method to measure AI bias more accurately, showing that cultural differences affect safety mechanisms in large language models. This could help create fairer AI systems worldwide.
Researchers have identified three main reasons why AI annotators disagree on safety policies. Understanding these differences can help improve AI safety guidelines. This matters because clearer policies mean safer AI for everyone.
Researchers propose a way to measure how much AI systems act with purpose, like a human. This could help us hold AI accountable for its actions. The framework defines intentionality as a set of behaviors, not consciousness, and shows how design choices affect these behaviors.
Researchers have created a new benchmark to test how AI agents handle information they can't access due to security restrictions. This helps ensure AI systems provide accurate responses without revealing sensitive data.
Researchers have developed a method to help AI personal assistants learn skills while keeping context limited. This could make local AI helpers smarter without sacrificing privacy or performance. The approach, called constant-context skill learning, allows AI agents to operate tools and browsers more efficiently, reducing the need for repeated processing of long histories.
Researchers have developed BALAR, a system that helps AI models actively seek missing information in conversations. This could make AI assistants much more effective at multi-step tasks.
Researchers developed FinAgent-RAG, an AI tool that helps answer complex questions about financial documents. It can handle tables, text, and footnotes to provide accurate insights.
Researchers have developed a new AI framework called LANTERN that helps machines learn new tasks faster by combining knowledge from multiple past experiences. This could make AI systems more efficient and adaptable in real-world applications.
Researchers used AI to automate the design of complex scientific formulas, potentially speeding up material science breakthroughs. This could lead to better batteries, stronger metals, and more efficient solar panels in the future.
Researchers found a way to bypass watermarks on AI-generated text, raising concerns about detecting fake content. This could make it harder to tell if text was written by a human or an AI.
Researchers have developed a way to make AI models better at admitting when they don't know something. This could make AI assistants more reliable in everyday use. The method works without needing to see the model's internal workings, making it useful for commercial AI services.
Researchers created a dataset to help AI work with medical tools. This could make AI more useful for doctors and scientists. The dataset, called BioTool, teaches AI to use specialized medical software and databases more effectively.
Researchers have identified a flaw in AI repair systems where rankings change unpredictably. They've released a tool to help developers spot and fix these issues. This could make AI systems more reliable for everyday users.
Researchers have developed a new way to watermark AI-generated text without reducing its quality. SLAM marks text structure rather than word choices, making it harder to detect and remove.
Researchers have developed a way to convert AI reasoning into symbolic solvers, making them faster and more accurate for complex programming tasks. This could lead to more efficient AI tools for everyday problem-solving.
Researchers created XL-SafetyBench to test AI models on country-specific safety issues and cultural sensitivities. This tool helps ensure AI understands and respects local norms beyond general English benchmarks.
Researchers have developed a way to create search-focused summaries from datasets that weren't originally designed for this purpose. This could make it easier to find key information in long documents without needing specialized data.
Researchers have developed a method to spot harmful intentions hidden in multi-turn AI conversations. This helps prevent AI models from being tricked into harmful behavior over time.
Researchers created a dataset to teach AI when to speak in group chats, preventing interruptions. This could make AI assistants more useful in meetings and group discussions.
New research shows that AI models handle negative emotions in early stages and positive ones later. This could help make AI responses more emotionally balanced and nuanced.
Researchers developed AdaGATE, a new method to help AI answer complex questions that require multiple steps. It improves accuracy by selecting the most relevant information and filling in gaps automatically.