Papers

Artificial Intelligence is revolutionizing healthcare, particularly in the realm of medical imaging. We now have Medical Large Vision Language Models (Med-LVLMs) capable of looking at an X-ray or a retinal scan and answering clinical questions. However, there is a persistent “elephant in the room” with these models: Hallucinations. Even the best models sometimes generate medical responses that sound plausible but are factually incorrect. In a clinical setting, a factual error isn’t just a glitch—it’s a safety risk. ...

Introduction We are living in the golden age of Large Language Models (LLMs). From GPT-4 to Llama, these models can write poetry, code, and essays with startling fluency. However, anyone who has spent time prompting these models knows they can be like unruly teenagers: talented but difficult to control. You might ask for a summary suitable for a five-year-old, and the model might return a text full of academic jargon. Worse, you might ask for a story, and the model might inadvertently produce toxic or biased content based on its training data. ...

If you follow the rapid evolution of Large Language Models (LLMs), you are likely familiar with the “alignment” phase. After a model consumes terabytes of text to learn how to predict the next token (Pre-training) and learns to follow instructions (Supervised Fine-Tuning or SFT), it undergoes a final polish: Preference Optimization. This is the stage where models like ChatGPT or Claude learn to be helpful, harmless, and conversational, usually via techniques like Reinforcement Learning from Human Feedback (RLHF) or Direct Preference Optimization (DPO). ...

Imagine you are asking a robot to help you in the kitchen. You say, “Pass me the red mug.” In a perfect world—or at least in the world of classic computer vision benchmarks—there is exactly one red mug on the counter. The robot identifies it, boxes it, and grabs it. But what if real life happens? What if the dishwasher is empty and there are no red mugs? Or what if you just had a party and there are three red mugs? ...

Introduction We often think of Large Language Models (LLMs) as vast repositories of knowledge, but they have a significant weakness: they cannot memorize everything, especially real-time events or niche domain knowledge. To solve this, the AI community widely adopted Retrieval-Augmented Generation (RAG). The concept is simple: when an LLM is asked a question, it first searches an external database (like Wikipedia) for relevant documents, then uses those documents to generate an answer. ...

Retrieval-Augmented Generation (RAG) has become the backbone of modern AI knowledge systems. By combining a parametric memory (the weights of a Large Language Model) with non-parametric memory (an external database of documents), we can build systems that answer questions with up-to-date, specific information. But anyone who has built a RAG system knows the dirty secret: Retrievers are noisy. If a user asks a question and the retriever fetches irrelevant documents, the generator (the LLM) is placed in a difficult position. It might try to force an answer from the bad context, leading to hallucinations, or it might ignore the context entirely, leading to answers that lack citation. Standard retrievers provide a similarity score, but this score is relative—it tells you that Document A is better than Document B, but it doesn’t tell you if Document A is actually good. ...

The rise of Text-to-Image (T2I) generation models like Stable Diffusion and Midjourney has revolutionized digital creativity. However, getting a high-quality image out of these models often requires “prompt engineering”—the art of crafting detailed, specific text descriptions. Because most users aren’t experts at writing these complex prompts, a new class of tools has emerged: Text-to-Image Prompt Refinement (T2I-Refine) services. These services take a user’s simple input (e.g., “a smart phone”) and expand it into a rich description (e.g., “a smart phone, intricate details, 8k resolution, cinematic lighting”). While helpful, this intermediate layer introduces a fascinating security and ethical question: Can a prompt refinement model be “poisoned” to secretly manipulate the output? ...

Introduction If you have ever asked ChatGPT or GitHub Copilot to write a Python script or a JavaScript function, you know the results can be magically accurate. These models have been trained on billions of lines of code from popular languages, making them incredibly proficient at standard programming tasks. But what happens when you step off the beaten path? When you ask an LLM to generate code for low-resource languages—domain-specific languages (DSLs) like Microsoft Power Query M, OfficeScript, or complex Excel formulas—the performance drops significantly. These languages don’t have the massive repositories of open-source code required to train a model effectively. ...

Beyond Wikipedia: Benchmarking Long-Form RAG with LFRQA and RAG-QA Arena Retrieval-Augmented Generation (RAG) has become the de facto architecture for building reliable AI systems. By connecting Large Language Models (LLMs) to external data sources, we give them a “memory” that is up-to-date and verifiable. However, a significant gap exists in how we evaluate these systems. Most current benchmarks rely on Wikipedia data and expect short, punchy answers (like “Paris” or “1984”). But in the real world, we use RAG to generate comprehensive reports, summarize financial trends, or explain complex biological mechanisms. When an LLM generates a nuanced, three-paragraph explanation, comparing it to a three-word ground truth using standard “exact match” metrics is like grading a history essay with a math answer key. ...

The release of Large Language Models (LLMs) like ChatGPT and LLaMA has fundamentally changed how we generate text. From writing emails to coding, the utility is undeniable. However, this power comes with a shadow: academic dishonesty, disinformation campaigns, and sophisticated phishing. To counter this, a new industry of “AI Detectors” has emerged—tools designed to distinguish between human and machine-written content. But how robust are these guardians? In this post, we dive deep into a paper titled “RAFT: Realistic Attacks to Fool Text Detectors,” which proposes a novel framework for “red-teaming” (attacking) these detectors. Unlike previous methods that often produce garbled or grammatically incorrect text, RAFT generates attacks that are essentially invisible to the human eye but confusing enough to break the best detectors available. ...

In the rapidly evolving world of Large Language Models (LLMs), engineers and researchers are constantly battling a frustrating trade-off: Speed versus Faithfulness. Imagine you are building a chatbot for a hotel. A user asks, “Do you have vegan breakfast options?” Option A: The bot answers instantly (0.5 seconds), but it hallucinates and says “Yes,” when you actually don’t. This is fast but dangerous. Option B: The bot pauses, retrieves the latest menu from a database, reads it, and answers correctly. This takes 3-5 seconds. This is accurate but feels sluggish to the user. Option A represents Retrieval Free Generation (RFG), where the model relies on its internal memory. Option B represents Retrieval Augmented Generation (RAG), the current gold standard for accuracy. ...

The world of Natural Language Processing (NLP) has been dominated by classical giants—Recurrent Neural Networks (RNNs), LSTMs, and more recently, Transformers. These models rely on massive computational resources and millions (sometimes billions) of parameters to understand human language. But a new paradigm is emerging: Quantum Computing. For years, the intersection of Quantum Computing and NLP—often called QNLP—was largely theoretical. However, as we approach the era of “quantum advantage,” researchers are asking a pivotal question: Can quantum models perform sequential tasks like text classification, and can they do it efficiently? ...

When we look at a photograph, we assume we are seeing objective reality. If you show a picture of a park to a person in New York and a person in Munich, surely they are seeing the same grass, the same benches, and the same sky. But are they? Cognitive science and psychology suggest that visual perception is deeply tied to culture. A Western viewer might focus intently on the foreground objects—the specific people or the specific make of a car. An East Asian viewer might place significantly more weight on the context, the background, and the relationships between objects. ...

The capabilities of Large Language Models (LLMs) have exploded in recent years, but they still have a critical limitation: they are confined to their training data. To solve real-world problems—like checking the weather, querying a database, or performing complex math—LLMs need to use external tools. Training a model to use tools (APIs) usually requires massive amounts of data containing instructions and the corresponding API calls. Because creating this data by hand is expensive, researchers rely heavily on synthetic data generated by other LLMs (like GPT-4). ...

Imagine you are trying to find your keys in a pitch-black room. You reach out and feel a smooth, cold surface. If you assume you are touching the kitchen counter, you might start feeling around for the fruit bowl. But if you were actually touching a glass table in the living room, every subsequent move you take is based on a false premise. You aren’t just lost; you are actively moving further away from your goal. ...

Introduction We often treat Large Language Models (LLMs) like omniscient reasoning engines. We feed them complex scenarios, legal documents, or medical summaries and ask them to derive conclusions. But how good are these models really when the answers aren’t black and white? In the real world, certainty is a luxury. A doctor doesn’t say “This symptom always means X.” They say, “Given this symptom, the diagnosis is highly likely to be X, unless Y is also present.” This is probabilistic reasoning—the ability to handle uncertainty, weigh evidence, and update beliefs. ...

Large Language Models (LLMs) like LLaMA-2 and Falcon have revolutionized AI, but they come with a hefty price tag: computational cost. Running these massive models requires significant GPU memory and processing power. To make these models accessible on local hardware or cheaper cloud instances, researchers have been racing to compress them. The most common technique is quantization—reducing the precision of the numbers used to represent the model. Instead of using 16-bit floating-point numbers (FP16), we try to squeeze them into 8 bits or even 4 bits. ...

When we read a news article or follow a complex narrative, we don’t process sentences in isolation. We instinctively look for connections. We ask ourselves: Why is the author saying this? What question does this sentence answer regarding the previous one? This cognitive process is the foundation of a linguistic framework known as Question Under Discussion (QUD). It treats discourse as a hierarchy of questions and answers. While humans do this naturally, teaching machines to parse these structures is notoriously difficult. Existing models often struggle to generate questions that make sense within the context or fail to link sentences correctly. ...

Imagine you are building an AI tutor designed to help students study for history exams. You feed the AI a textbook chapter about the Mongol Empire and ask it to generate quiz questions. The AI outputs: “Who was the name of Ögedei’s wife?” Grammatically, it’s a bit clunky. But technically, if you look at the source text, the words match. Traditional evaluation metrics might give this a decent score. However, a human student would find it confusing. ...

The rise of Large Language Models (LLMs) like GPT-4 and LLaMA has been nothing short of meteoric. We have seen them pass bar exams, write code, and compose poetry. However, there is a distinct difference between generating fluent text and performing rigorous logical reasoning. While an LLM can explain the history of chess, can it actually play chess at a high level? Can it solve a cryptic crossword or deduce the location of a mine in Minesweeper? ...