Quick keys
— Navigate with ← → or spacebar.
F
fullscreen ·
S
speaker view ·
O
overview ·
Alt+Click
zoom
Navigation
×
# Day 1: State of Large Language Models <div style="margin-top: 3em; padding: 0 2em; text-align: center;"> <div style="margin-bottom: 1em;"> SCIPE Workshop on Large Language Models </div> <div style="opacity: 0.8; font-size: 0.9em;"> Case Western Reserve University<br> <span style="opacity: 0.7;">January 16, 2026</span> </div> </div> <div style="text-align: center; margin-top: 3em;"> <div style="display: flex; align-items: center; justify-content: center; gap: 3rem; flex-wrap: wrap;"> <img src="/assets/slides/2026-01-16-state-of-llms/logo/cwru_logo.webp" alt="CWRU" style="height: 90px; width: auto; padding: 0.4em 0.6em;" /> <img src="/assets/slides/2026-01-16-state-of-llms/logo/uoc_logo.webp" alt="University of Cincinnati" style="height: 90px; width: auto; padding: 0.4em 0.6em;" /> <img src="/assets/slides/2026-01-16-state-of-llms/logo/osu_logo.webp" alt="Ohio State University" style="height: 90px; width: auto; padding: 0.4em 0.6em;" /> <img src="/assets/slides/2026-01-16-state-of-llms/logo/NSF_Official_logo_Low_Res_150ppi-300x300.webp" alt="NSF" style="height: 90px; width: auto; padding: 0.4em 0.6em;" /> </div> </div> <!-- --- ## Table of Contents <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 3rem;"> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem; border-radius: 12px; text-align: left; border-left: 4px solid var(--r-main-color);"> <h3 style="margin: 0 0 0.5rem 0; font-size: 1.1em; color: var(--r-heading-color);">1. [Placeholder] Foundations</h3> <div style="font-size: 0.75em; opacity: 0.8;">What an LLM is, how it’s trained, why scaling helps</div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem; border-radius: 12px; text-align: left; border-left: 4px solid var(--r-main-color);"> <h3 style="margin: 0 0 0.5rem 0; font-size: 1.1em; color: var(--r-heading-color);">2. [Placeholder] Language</h3> <div style="font-size: 0.75em; opacity: 0.8;">Tokens, meaning, and what “learning language” might mean</div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem; border-radius: 12px; text-align: left; border-left: 4px solid var(--r-main-color);"> <h3 style="margin: 0 0 0.5rem 0; font-size: 1.1em; color: var(--r-heading-color);">3. [Placeholder] Reasoning & Tools</h3> <div style="font-size: 0.75em; opacity: 0.8;">In-context learning, planning, and tool use</div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem; border-radius: 12px; text-align: left; border-left: 4px solid var(--r-main-color);"> <h3 style="margin: 0 0 0.5rem 0; font-size: 1.1em; color: var(--r-heading-color);">4. [Placeholder] Limits & Open Questions</h3> <div style="font-size: 0.75em; opacity: 0.8;">Hallucinations, evaluation, alignment, and “stochastic parrots”</div> </div> </div> --> --- ## Workshop Overview <div style="display: grid; grid-template-columns: 1fr 1fr 1fr; gap: 3rem; margin-top: 2rem;"> <div style="background: linear-gradient(135deg, rgba(66, 185, 131, 0.2) 0%, rgba(66, 185, 131, 0.05) 100%); padding: 2rem; border-radius: 16px; border-left: 5px solid rgba(66, 185, 131, 0.8); box-shadow: 0 4px 15px rgba(0,0,0,0.1);"> <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1.5rem;"> <div style="font-size: 3em; font-weight: bold; color: rgba(66, 185, 131, 1);"> Day 1 </div> <!-- <div style="font-size: 1.3em; font-weight: bold; letter-spacing: 0.5px; color: var(--r-heading-color);">Day 1</div> --> </div> <div style="font-size: 0.85em; line-height: 1.8; text-align: left;"> <div style="margin-bottom: 1.2rem; padding-left: 0.5rem; border-left: 2px solid rgba(66, 185, 131, 0.4);"> <strong style="color: rgba(66, 185, 131, 1);">Introduction to Large Language Models</strong> </div> <ul style="list-style: none; padding: 0; margin: 0;"> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(66, 185, 131, 1);">▸</span> LLM in Action: Commercial Models </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(66, 185, 131, 1);">▸</span> Open-Source LLMs </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(66, 185, 131, 1);">▸</span> Inference & Efficiency </li> </ul> </div> </div> <div style="background: linear-gradient(135deg, rgba(255, 99, 71, 0.2) 0%, rgba(255, 99, 71, 0.05) 100%); padding: 2rem; border-radius: 16px; border-left: 5px solid rgba(255, 99, 71, 0.8); box-shadow: 0 4px 15px rgba(0,0,0,0.1);"> <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1.5rem;"> <div style="font-size: 3em; font-weight: bold; color: rgba(255, 99, 71, 1);"> Day 2 </div> <!-- <div style="font-size: 1.3em; font-weight: bold; letter-spacing: 0.5px; color: var(--r-heading-color);">Day 2</div> --> </div> <div style="font-size: 0.85em; line-height: 1.8; text-align: left;"> <div style="margin-bottom: 1.2rem; padding-left: 0.5rem; border-left: 2px solid rgba(255, 99, 71, 0.4);"> <strong style="color: rgba(255, 99, 71, 1);">Advanced Applications & Techniques</strong> </div> <ul style="list-style: none; padding: 0; margin: 0;"> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(255, 99, 71, 1);">▸</span> Tool Use & Integration </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(255, 99, 71, 1);">▸</span> Retrieval-Augmented Generation (RAG) </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(255, 99, 71, 1);">▸</span> Hands-on Coding Session </li> </ul> </div> </div> <div style="background: linear-gradient(135deg, rgba(30, 144, 255, 0.2) 0%, rgba(30, 144, 255, 0.05) 100%); padding: 2rem; border-radius: 16px; border-left: 5px solid rgba(30, 144, 255, 0.8); box-shadow: 0 4px 15px rgba(0,0,0,0.1);"> <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1.5rem;"> <div style="font-size: 3em; font-weight: bold; color: rgba(30, 144, 255, 1);"> Day 3 </div> <!-- <div style="font-size: 1.3em; font-weight: bold; letter-spacing: 0.5px; color: var(--r-heading-color);">Day 3</div> --> </div> <div style="font-size: 0.85em; line-height: 1.8; text-align: left;"> <div style="margin-bottom: 1.2rem; padding-left: 0.5rem; border-left: 2px solid rgba(30, 144, 255, 0.4);"> <strong style="color: rgba(30, 144, 255, 1);">Research Areas in LLMs</strong> </div> <ul style="list-style: none; padding: 0; margin: 0;"> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(30, 144, 255, 1);">▸</span> Attention, under the hood </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(30, 144, 255, 1);">▸</span> LLM Research Roadmap </li> <li style="margin-bottom: 0.8rem; padding-left: 1.5rem; position: relative;"> <span style="position: absolute; left: 0; color: rgba(30, 144, 255, 1);">▸</span> Paper Discussion and Research Ideas </li> </ul> </div> </div> </div> <div style="text-align: center; margin-top: 2.5rem; font-size: 0.8em; opacity: 0.85; font-style: italic;"> <div style="background: rgba(255, 255, 255, 0.05); padding: 1rem 2rem; border-radius: 8px; display: inline-block;"> <!-- <strong>From Theory to Practice:</strong> Building production-ready LLM applications --> </div> </div> --- ## Strengthening the Cyberinfrastructure Professionals Ecosystem (SCIPE) <div style="display: grid; grid-template-columns: 1.2fr 0.8fr; gap: 2rem; margin-top: 1rem;"> <div style="font-size: 0.8em; text-align: left;"> **OH-SCIPE** is an NSF-funded initiative building an interdisciplinary research support community for AI and data sciences across Ohio institutions. <div style="margin-top: 1.5rem;"> **Partner Institutions:** - Case Western Reserve University - University of Cincinnati - Ohio Supercomputer Center </div> <!-- <div style="margin-top: 1.5rem; background: rgba(255, 255, 255, 0.05); padding: 1rem; border-radius: 8px; border-left: 3px solid var(--r-main-color);"> <strong>Mission:</strong> Increase awareness of opportunities for using and supporting AI as a <strong>Cyberinfrastructure Professional (CIP)</strong> – individuals who provide IT support to university researchers with a growing focus on AI. </div> --> </div> <div style="display: flex; flex-direction: column; align-items: center; justify-content: center; gap: 1.5rem;"> <div style="width: 100%; background: rgba(255,255,255,0.06); padding: 1.2rem; border-radius: 12px; text-align: center; font-size: 0.75em;"> <img src="/assets/slides/2026-01-16-state-of-llms/logo/NSF_Official_logo_Low_Res_150ppi-300x300.webp" alt="NSF" style="height: 100px; width: auto; margin-bottom: 0.8rem;" /> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem;">NSF Awards</div> <div style="font-size: 1.4em; font-weight: bold; color: var(--r-main-color);">$5.1 Million</div> <div style="margin-top: 0.5rem; opacity: 0.8;">Total funding for the OH-SCIPE project</div> </div> <div style="text-align: center; margin-top: 0.5rem;"> <div style="display: flex; align-items: center; justify-content: center; gap: 2rem; flex-wrap: wrap;"> <img src="/assets/slides/2026-01-16-state-of-llms/logo/cwru_logo.webp" alt="CWRU" style="height: 60px; width: auto;" /> <img src="/assets/slides/2026-01-16-state-of-llms/logo/uoc_logo.webp" alt="UC" style="height: 60px; width: auto;" /> <img src="/assets/slides/2026-01-16-state-of-llms/logo/osu_logo.webp" alt="OSU" style="height: 60px; width: auto;" /> </div> </div> </div> </div> --- ## OH-SCIPE Projects and Background <div style="font-size: 0.8em; text-align: left;"> <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem;"> <div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.2rem; border-radius: 12px; margin-bottom: 1.2rem; border-left: 3px solid var(--r-main-color);"> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem; font-size: 0.9em;">Mission</div> <div style="line-height: 1.8;"> <strong>Revolutionizing Research with AI and ML Integration</strong> <div style="margin-top: 0.8rem; opacity: 0.9;"> AI/ML technologies have significantly accelerated research across all fields of Science and Engineering. OH-SCIPE aims to further integrate cutting-edge machine learning engineering into S&E research initiatives. </div> </div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.2rem; border-radius: 12px;"> <!-- <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem; font-size: 0.9em;">Legacy of CI Support</div> --> <div style="line-height: 1.8;"> <!-- Building on the foundation established by the <strong>Coalition for Academic Scientific Computation (CASC)</strong> in 1989, OH-SCIPE is dedicated to enhancing S&E research by facilitating the incorporation of <strong>ML Engineers</strong> into research teams. --> </div> </div> </div> <div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.2rem; border-radius: 12px; margin-bottom: 1.2rem;"> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem; font-size: 0.9em;">Current Initiatives</div> <ul style="margin: 0; padding-left: 1.2em; line-height: 1.8;"> <li><strong>SCIPE Workshop on Large Language Models</strong> (January 2026)</li> <li><strong>Summer 2026 Internship</strong> - AI Research Experience</li> <li><strong>Research Project Support</strong> - ML engineering integration</li> <li><strong>Training Programs</strong> - AI, HPC, and computational tools</li> </ul> </div> <div style="background: rgba(66, 185, 131, 0.15); padding: 1.2rem; border-radius: 12px; border-left: 3px solid rgba(66, 185, 131, 0.8);"> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.9; margin-bottom: 0.6rem; font-size: 0.9em; font-weight: bold;">Connect With Us</div> <div style="line-height: 1.6;"> Explore how OH-SCIPE can transform your research. Partner with an ML Engineer who can take your project to the next level. </div> <div style="margin-top: 0.8rem; opacity: 0.8; font-size: 0.95em;"> 🌐 <a href="https://oh-scipe.github.io/" style="color: inherit;">oh-scipe.github.io</a> </div> </div> </div> </div> </div> --- ## Team <div style="font-size: 0.9em; max-width: 75rem; margin: 2rem auto;"> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem 2rem; border-radius: 12px; margin-bottom: 1.2rem; border-left: 3px solid var(--r-main-color);"> <div style="display: grid; grid-template-columns: auto 1fr; gap: 1rem 2rem; align-items: center;"> <div style="font-weight: bold; color: var(--r-heading-color); font-size: 1.1em;">PIs:</div> <div style="line-height: 1.6;">Vipin Chaudhary, Karen Tomko, Ketaki Mhaisekar</div> </div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem 2rem; border-radius: 12px; margin-bottom: 1.2rem; border-left: 3px solid var(--r-main-color);"> <div style="display: grid; grid-template-columns: auto 1fr; gap: 1rem 2rem; align-items: center;"> <div style="font-weight: bold; color: var(--r-heading-color); font-size: 1.1em;">CIPs:</div> <div style="line-height: 1.6;">Mohsen Hariri, Weicong Chen, Evan Jaffe, Tony Lamping</div> </div> </div> <div style="background: rgba(255, 255, 255, 0.05); padding: 1.5rem 2rem; border-radius: 12px; border-left: 3px solid var(--r-main-color);"> <div style="display: grid; grid-template-columns: auto 1fr; gap: 1rem 2rem; align-items: start;"> <div style="font-weight: bold; color: var(--r-heading-color); font-size: 1.1em; padding-top: 0.2rem;">Volunteer Students:</div> <div style="line-height: 1.6;">Cheng Guo, Vikash Singh, Gabriel Natenshon, Wang Yang, Shouren Wang, Nahal Shahini, Nat Tomczak, Kutay Tasci</div> </div> </div> </div> --- ## Language: What is the object we’re modeling? <div style="display: grid; grid-template-columns: 1.2fr 0.8fr; gap: 2rem; margin-top: 1rem;"> <div style="font-size: 0.8em; text-align: left;"> - **Language ≠ text**: text is a *record*; language is a *system* behind it. - **Semantics + pragmatics**: meaning depends on context, goals, and shared assumptions. - **Compositionality**: we build new meanings by combining old parts. - **Grounding**: words point to the world (perception, action, social reference). <div style="margin-top: 1.2rem; opacity: 0.85;"> Working view for this talk: LLMs are trained on <strong>E-language artifacts</strong> (text), and we ask what they infer about the <strong>latent structure</strong> that produced them. </div> </div> <div style="display: flex; align-items: center; justify-content: center;"> <div style="width: 100%; background: rgba(255,255,255,0.06); padding: 1.2rem; border-radius: 12px; text-align: left; font-size: 0.75em;"> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem;">Key Distinction</div> <div style="margin-bottom: 0.8rem;"><strong>I-language</strong>: internal competence (mental grammar)</div> <div><strong>E-language</strong>: externalized utterances (texts, corpora)</div> </div> </div> </div> --- ## Language in LLMs: from tokens to meaning <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem;"> <div style="font-size: 0.8em; text-align: left;"> - Training objective (causal LM): $p(x) = \\prod_t p(x_t \\mid x_{<t})$ - The model sees <strong>token sequences</strong>, not intentions, speakers, or the world. - Yet it can learn <strong>latent variables</strong> that help prediction (topics, entities, relations). - Meaning can appear as <strong>use</strong>: representations that support accurate continuation and task performance. </div> <div style="font-size: 0.75em; text-align: left;"> <div style="background: rgba(255,255,255,0.06); padding: 1.2rem; border-radius: 12px;"> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.8rem;">Three “Levels”</div> 1) <strong>Form</strong>: tokens, syntax, style<br> 2) <strong>Content</strong>: facts, concepts, relations<br> 3) <strong>Use</strong>: goals, planning, tool calls <div style="margin-top: 0.9rem; opacity: 0.8;"> Big question: does optimizing next-token prediction inevitably induce <em>something like</em> a world model? </div> </div> </div> </div> --- ### Language and Thought <img src="/assets/slides/2025-10-27-psychometric-modeling/image-6.webp" alt="Alt text" style="width:50%; height:auto; display:block; margin-left:auto; margin-right:auto;"> <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem; font-size: 0.72em; text-align: left;"> <div> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.6rem;">Two Claims People Often Mix</div> - <strong>Thought requires language</strong> (strong linguistic determinism) - <strong>Language shapes thought</strong> (weaker, more plausible) <div style="margin-top: 0.8rem; opacity: 0.85;"> Even if there is “mentalese,” language is a powerful <strong>interface</strong>: it compresses, externalizes, and coordinates cognition socially. </div> </div> <div> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.6rem;">So what about LLMs?</div> - They model <strong>E-language</strong> extremely well (distribution over utterances). - If internal states act like a compact latent model, they may behave as if they have <strong>concepts</strong>. - But <strong>human-like I-language</strong> is not guaranteed: grounding, embodiment, memory, and goals differ. <div style="margin-top: 0.8rem; opacity: 0.85;"> <!-- Useful question: what kinds of “thought-like” structure are <em>necessary</em> for predictive language competence? --> </div> </div> </div> --- ## Are they human-like thinkers or just stochastic parrots? <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem; font-size: 0.75em; text-align: left;"> <div> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.6rem;">Why it looks like thinking</div> - Solves novel tasks with <strong>instructions + examples</strong> (in-context learning). - Shows <strong>systematic generalization</strong> in some domains (math, code, logic). - Can <strong>plan and verify</strong> when scaffolded (tools, self-checks, search). </div> <div> <div style="text-transform: uppercase; letter-spacing: 1px; opacity: 0.7; margin-bottom: 0.6rem;">Why the “parrot” critique persists</div> - Sensitive to prompts; can be <strong>brittle</strong> under distribution shift. - Produces fluent but wrong answers (<strong>hallucinations</strong>). - Can rely on <strong>surface heuristics</strong> rather than causal understanding. </div> </div> <img src="/assets/slides/2025-10-27-psychometric-modeling/image-16.webp" alt="Alt text" style="width:60%; height:auto; display:block; margin-left:auto; margin-right:auto;"> --- ## LLM Use Cases <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem; align-items: center;"> <div style="font-size: 0.85em; text-align: left;"> * Conversational AI & chatbots * Code generation & software development assistance * Text summarization * Information retrieval & question answering * Content creation & marketing * Education & personalized tutoring * Healthcare documentation & clinical support * Legal & compliance analysis * Translation & multilingual communication * Data analysis & business intelligence </div> <div style="display: flex; align-items: center; justify-content: center;"> <img src="/assets/slides/2026-01-16-state-of-llms/image-3.webp" alt="LLM Use Cases" style="width: 70%; height: auto; max-width: 90%;" /> </div> </div> --- ## Agentic LLMs <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 2rem; margin-top: 1rem; align-items: center;"> <div style="font-size: 0.85em; text-align: left;"> **Agents are models using tools in a loop** <div style="margin-top: 1.5rem;"> **Common Tools:** - Search APIs (e.g., Bing, Google) - Compilers - Use web browsers - Calculators - File system access - External knowledge bases </div> </div> <div style="display: flex; align-items: center; justify-content: center;"> <img src="/assets/slides/2026-01-16-state-of-llms/image.webp" alt="Agentic LLMs" style="width: 100%; height: auto; max-width: 90%;" /> </div> </div> --- ## Roadmap to LLMs: Big Picture  --- # Thank You! <div style="display: flex; flex-direction: column; align-items: center; justify-content: center; margin-top: 4rem;"> <h3 style="margin-bottom: 2rem; font-size: 2em;">Questions?</h3> <div style="text-align: center;"> <img src="/assets/slides/2026-01-16-state-of-llms/SCIPE_WORKSHOP.webp" alt="SCIPE Workshop QR Code" style="width: 27rem; max-width: 100%; height: auto; margin: 2rem auto;" /> </div> <div style="font-size: 1.5em; font-weight: bold; text-align: center; margin-top: 2rem; opacity: 0.95;"> Now let's start coding! </div> </div> </div>