Large language models (LLMs) have gained significant attention in recent years, but ensuring their safe and ethical use remains a critical challenge. Researchers are focused on developing effective alignment procedures to calibrate these models to adhere to human values and safely follow human intentions. The primary goal is to prevent LLMs from engaging in unsafe… →

The field of research focuses on optimizing algorithms for training large language models (LLMs), which are essential for understanding and generating human language. These models are critical for various applications, including natural language processing and artificial intelligence. Training LLMs requires significant computational resources and memory, making optimizing these processes a high-priority area for researchers. The… →

Frontier AI systems, including LLMs, increasingly shape human beliefs and values by serving as personal assistants, educators, and authors. These systems, trained on vast amounts of human data, often reflect and propagate existing societal biases. This phenomenon, known as value lock-in, can entrench misguided moral beliefs and practices on a societal scale, potentially reinforcing problematic… →

Upon scanning their code for vulnerabilities, companies frequently encounter numerous findings. It takes an average of three months for firms to resolve a vulnerability, and 60% of those breached knew about the unpatched vulnerability used. Engineers tend to focus less on security patches in favor of work that generates cash. Fixing vulnerabilities is extremely costly… →

The rise of Generative AI (GenAI) has revolutionized various industries, from healthcare and finance to entertainment and customer service. The effectiveness of GenAI systems hinges on the seamless integration of four critical components: Human, Interface, Data, and large language models (LLMs). Understanding these elements is essential for designing robust and efficient GenAI workflows. Human Humans… →

Large Language Models (LLMs) have shown impressive performance in a range of tasks in recent years, especially classification tasks. These models demonstrate amazing performance when given gold labels or options that include the right answer. A significant limitation is that if these gold labels are purposefully left out, LLMs would still choose among the possibilities,… →

Multi-modal Large Language Models (MLLMs) have various applications in visual tasks. MLLMs rely on the visual features extracted from an image to understand its content. When a low-resolution image containing fewer pixels is provided as input, it translates less information to these models to work with. Due to this limitation, these models often need to… →