LLMs & Generative AI
← Back to Artificial IntelligenceLarge language models, text generation, multimodal models, and prompt engineering.
This section explores the structural mechanics, scaling limits, and real-world applications of large language models and generative architectures. Our research investigates the entire lifecycle of these systems, beginning with the transition from pre-training and supervised fine-tuning to post-training alignment paradigms like direct preference optimization. We dissect the fundamental constraints of transformer architectures, examining how subword tokenization algorithms introduce cognitive blind spots—such as the "strawberry problem"—and how long-context systems optimize memory bandwidth through techniques like speculative decoding, latent attention, and key-value cache management.
Beyond core training, this research analyzes the evolution of prompt engineering from basic linguistic prompting to structured systems architecture, including agentic workflows, chain-of-thought processing, and advanced reasoning topologies. These methodologies are weighed against the inherent limitations of next-token prediction, detailing why AI models exhibit hallucinations, fail at analogical reasoning, and suffer from biases incentivized by flawed evaluation benchmarks. This structural analysis extends into multimodality, mapping how latent diffusion models, generative adversarial networks, and early-fusion architectures enable image, video, music, and de novo protein generation.
In practical application, we evaluate how these models perform in high-stakes environments. This includes the implementation of LLMs in quantitative finance—covering alpha generation, trade execution latency, and secure fine-tuning on proprietary data using parameter-efficient methods and differential privacy—as well as their utility in legal reasoning. Finally, our research addresses the cognitive and societal footprint of generative technology. We examine the tension between human learning and automated assistance, exploring how cognitive offloading and epistemic debt alter human writing skills, educational assessment frameworks, and long-term creative capacity.
40 published articles
- Large language models for trading and alpha generation Discover how large language models generate financial alpha, parse corporate disclosures, and face critical look-ahead bias in backtesting. 2026-06-06
- Can You Use ChatGPT for Swing Trading Learn how swing traders use ChatGPT for sentiment analysis and coding while avoiding critical math and risk management failures. 2026-06-06
- Why AI Models Are So Confidently Wrong Learn why large language models generate confident wrong answers and how evaluation benchmarks incentivize AI hallucinations over admitting uncertainty. 2026-05-31
- What Is Tokenization and How AI Reads Your Text Discover how AI processes text through tokenization, the subword algorithms like BPE, and why it causes cognitive blind spots like the strawberry problem. 2026-05-31
- What Fits in an AI Context Window and What It Costs Learn how tokenization, context caching, and architectural limits like the lost-in-the-middle effect impact the economics of AI context windows. 2026-05-31
- Is Prompt Engineering Still a Real Skill in 2026 Explore how prompt engineering in 2026 has evolved from linguistic hacks to systems architecture, context engineering, and agentic workflows. 2026-05-31
- How LLM Hallucinations Actually Work Learn why large language models hallucinate due to mathematical limits of next-token prediction, transformer architecture, and flawed evaluation benchmarks. 2026-05-31
- How Large Language Models Are Trained Learn how large language models are trained through pre-training, supervised fine-tuning, and human-preference alignment to become useful AI assistants. 2026-05-31
- How AI Generates Images Step by Step Learn how AI image generators work step-by-step, using diffusion models, text encoders, and latent space to turn prompts into realistic pictures. 2026-05-31
- A Guide to Large Language Models for Non-Engineers Understand how large language models work, including tokenization, the transformer architecture, training lifecycles, and scaling limits. 2026-05-31
- How Deepfakes Work, How to Spot Them, and Why They Matter Learn how deepfakes are created using GANs, how to spot synthetic media, and the global security risks of AI-generated audio and video. 2026-05-30
- Secure large language model fine-tuning on proprietary trading data Learn how quantitative finance firms securely fine-tune large language models on proprietary trading data using PEFT, RAG, and differential privacy. 2026-05-21
- Linguistic relativity in large language models Explore how large language models reveal insights into the Sapir-Whorf hypothesis of linguistic relativity and the symbol grounding problem. 2026-05-21
- Latency constraints on large language models in trade execution Explore how latency constraints and hardware bottlenecks limit the use of large language models for real-time trade execution in financial markets. 2026-05-21
- Large language models and the structure of legal reasoning This study examines LLMs in legal reasoning, showing why generative AI excels at rule extraction but fails at analogical reasoning and jurisprudence. 2026-05-21
- Effects of Chain-of-Thought Prompting on LLM Macro Trading Theses Explore how Chain-of-Thought prompting in LLMs impacts macro trading theses, detailing logic accuracy improvements and structural hallucination risks. 2026-05-21
- Balancing artificial intelligence and deep academic thinking Discover how to balance artificial intelligence with deep academic thinking by understanding cognitive offloading, epistemic debt, and slow productivity. 2026-05-20
- Socratic AI tutoring and conceptual understanding Socratic AI tutors foster deep learning by using conversational dialogue to prompt reflection, manage cognitive load, and reduce cognitive offloading. 2026-05-17
- Generative AI and New Models of Educational Assessment This research analyzes how generative AI transforms educational assessment by shifting from product-oriented evaluation to process-focused learning models. 2026-05-17
- AI and human feedback in writing, coding, and creative work Compare AI-generated feedback to human instructors in education and discover why hybrid models yield superior learning outcomes in writing and coding. 2026-05-17
- Text-to-video artificial intelligence technology and limitations This study examines text-to-video AI architectures like Sora, focusing on diffusion transformers, spacetime patches, and the limits of physical simulation. 2026-05-13
- Systematic Findings in Large Language Model Prompt Engineering Discover systematic findings in LLM prompting, including few-shot learning, temperature settings, and reasoning topologies like Chain and Graph of Thoughts. 2026-05-13
- Speculative decoding for large language models Speculative decoding accelerates LLM inference by 2-3x through parallel verification and draft models, overcoming memory bandwidth limits without quality loss. 2026-05-13
- Multimodal Reasoning in Artificial Intelligence Explore multimodal AI reasoning architectures, from late-fusion adapters to native models, and the cognitive vulnerabilities exposed by the Clever Hans effect. 2026-05-13
- Mechanisms of Large Language Model Hallucinations Explore the mathematical mechanisms behind AI hallucinations, including the impact of context window scaling and sparse Mixture-of-Experts architectures. 2026-05-13
- Mechanisms of in-context learning in large language models Explore how large language models use induction heads, implicit gradient descent, and task vectors to perform in-context learning without weight updates. 2026-05-13
- Latent Diffusion Architecture for AI Image Generation Explore the evolution of Stable Diffusion architecture, from VAE latent compression and U-Net backbones to modern Multimodal Diffusion Transformers. 2026-05-13
- Internal world models in artificial intelligence This research explores how AI systems like large language models and reinforcement learning agents build internal world models to simulate and predict reality. 2026-05-13
- Instruction tuning and post-training in large language models Learn how instruction tuning and post-training paradigms like DPO align large language models with human intent to unlock practical LLM usability. 2026-05-13
- Impact of Tokenization on Large Language Models Explore how subword tokenization defines the limits of LLM performance, from the strawberry problem to the economic impact of token fertility across scripts. 2026-05-13
- Engineering and Science of Long-Context Language Models Explore the engineering of 1M-token context windows, covering KV cache optimization, Multi-head Latent Attention, and distributed system scaling techniques. 2026-05-13
- Computational mechanisms of artificial intelligence reasoning Explore how AI reasoning works through test-time computation, chain-of-thought mechanisms, and transformer circuit complexity in large language models. 2026-05-13
- Chain-of-thought prompting in large language models Chain-of-thought prompting allows large language models to break down complex tasks into sequential reasoning steps for improved logical and mathematical acc... 2026-05-13
- Architecture of Multimodal Vision and Language Models Learn how multimodal AI architectures like GPT-4o and Gemini integrate vision and language using tokenization, projection, and early fusion strategies. 2026-05-13
- Architecture and Creative Questions in AI Music Generation Analyze AI music generation architecture, comparing Suno and Udio while exploring autoregressive models, latent diffusion, and modern production workflows. 2026-05-13
- AI diffusion models for protein design Learn how generative AI and diffusion models enable de novo protein design by engineering novel amino acid sequences for therapeutics and industrial use. 2026-05-13
- Longitudinal impacts of language model assistance on writing skill Longitudinal research explores how language models impact writing skills, revealing a dichotomy between productivity gains and long-term cognitive atrophy. 2026-05-12
- Language acquisition in humans and large language models This research compares human language acquisition and LLMs, highlighting differences in data efficiency, social grounding, and morphological learning. 2026-05-12
- The impact of artificial intelligence on human creativity Research examines how generative AI impacts human creativity through historical parallels, labor market shifts, and neuro-cognitive effects like cognitive at... 2026-05-11
- Cognitive offloading and generative artificial intelligence in 2026 Research examines how generative AI causes cognitive debt and reduced neural connectivity by shifting offloading from memory to executive delegation. 2026-05-11