AI Agents Struggle to Put Users First, Microsoft Study Finds

AI Delegation Flaws Exposed in Document Corruption Study

A new study reveals that large language models (LLMs) often corrupt documents when used for delegated tasks like document editing. Researchers tested 19 LLMs across 52 professional domains, including coding and music notation, and found that even advanced models-such as Gemini, Claude, and GPT-degraded content by an average of 25% in long workflows. This degradation worsened with larger documents, longer interactions, or the presence of distracting files. The study highlights a critical reliability issue in AI delegation, where errors silently compound over time, raising concerns about trustworthiness in professional settings. As AI adoption grows, addressing these flaws will be essential for maintaining accuracy and integrity in knowledge work.

AI Solves Complex Math Problems in Seconds

Recent advancements in large language models (LLMs) have shown they can tackle research-level math problems with remarkable speed. ChatGPT 5.5 Pro, for instance, solved a PhD-level problem in just an hour without needing any input beyond the question itself. This breakthrough comes after LLMs successfully solved several Erdős problems, initially thought to be too challenging for AI. While some solutions relied on existing literature, others demonstrated the ability to spot gaps in human knowledge. Now, mathematicians are realizing that if a problem has an easy solution humans missed, LLMs can find it. This raises the bar for creating new math challenges-problems must now be difficult enough to stump even the most advanced AI. As a result, researchers like Mel Nathanson are rethinking how they pose questions, ensuring they're tough enough for both humans and AI to grapple with. The future of mathematical exploration is likely to involve more collaboration between human intuition and machine efficiency.

AI Breakthrough Reduces Hallucinations in Vision-Language Models

Researchers have developed a new method called Positive-and-Negative Decoding (PND) that addresses a major issue with vision-language models (VLMs), which often generate incorrect or misleading content by relying too much on text-based assumptions. VLMs, like other AI systems, sometimes "hallucinate" objects in images because they prioritize language over visual data. This can lead to errors in tasks such as image captioning or object recognition. The PND framework works during the inference phase-meaning it doesn’t require retraining the model-and actively corrects this imbalance by giving more weight to visual evidence. It uses two pathways: one that emphasizes what should be present in the image and another that highlights what shouldn’t, creating a contrast that steers the AI toward more accurate results. Tests on datasets like POPE, MME, and CHAIR show significant improvements without needing additional training data or fine-tuning. This advancement is particularly important for industries relying on VLMs, such as robotics, healthcare imaging, and autonomous vehicles, where accuracy is critical. Developers can now trust these models to produce more reliable visual-grounded outputs. As AI continues to evolve, techniques like PND will likely become standard tools for ensuring the integrity of multimodal systems.

New AI Model Revolutionizes Antibody Design for Better Drug Development

Scientists have developed a groundbreaking artificial intelligence model that significantly improves the design of antibodies, which are crucial for treating diseases. Current methods struggle with creating diverse and effective antibody sequences, but this new approach uses advanced machine learning techniques to overcome these limitations. By focusing on the biological roots of antibody formation and using a novel "germline absorbing diffusion" method, the AI model reduces bias and enhances accuracy in predicting non-germline residues from 26% to 46%. This advancement is particularly important for drug developers aiming to create treatments that are both effective and stable. The model's ability to generate antibodies with improved hydrophobicity and binding affinity could lead to more efficient therapies. Researchers are already testing its potential in real-world applications, expecting it to accelerate the discovery of new medicines. As this technology evolves, experts predict it will become a vital tool for pharmaceutical companies, potentially reducing the time and cost associated with developing life-saving treatments. The future of AI in medicine looks promising, with this breakthrough paving the way for even more innovative solutions.

AI Models Struggle to Accurately Specify System Code

Researchers tested large language models on a benchmark called SysMoBench. The test checks how well these models can create accurate specifications for system code. The models did well on basic checks but struggled with more complex tests. They could compile and run the code, but often failed to accurately model the system. This matters because accurate specifications are crucial for ensuring system safety and reliability. For example, the models were given 11 systems to specify, including concurrent synchronization and distributed protocols. The results show that current models are not yet reliable for specifying system code. They can recall textbook examples, but struggle to abstract logic from complex implementations. Next, researchers will work to improve the models and make them more accurate.