Jordan Boyd-Graber

Почему использование простой языковой модели может ускорить работу более умной: спекулятивное дек...

Кэширование KV: ускорение вывода LLM [Лекция]

Нам не нужны никакие вонючие числа с плавающей точкой для ИИ [Лекция]

Оптимизация моделей: тонкая настройка, дистилляция, LoRA и QLoRA [Лекция]

Использование DSPy для оптимизации подсказок в Python: пример калибровки вопросов для Quiz Bowl [...

Почему я не сторонник подсказок для ИИ [Лекция]

Хороший план трудно найти: ИИ не может распознать, что помогает пользователю [Исследование]

Как* G*PT* превращает* символы* в* «*слова*»:* узнайте*,* что* означают* звёздочки* [Лекция]

Токенизация: как компьютеры переходят от битов к словам [Лекция]

Адам / AdamW: Чем оптимизатор «по умолчанию» отличается от SGD [Лекция]

AI is great at answering some questions, but what happens with bad questions? [Research, ACL 2025]

Why outsourcing document exploration to AI doesn't always work [ACL 2025, Research]

When AIs play trivia games, they don't know when they're wrong [ACL 2025, Research]

How to "backsolve" LLM personalization by generating user personas [ACL 2025, Research]

ProxAnn (Sting Parody): An AI Alternative to NPMI Evaluation of Topic Models [ACL 2025, Research]

How AI can detect offers that are "too good to be true" [ACL Findings 2025, Research]

The bad habit from high school that's making AI dumb [ACL 2025, Research]

We built an AI to help Noobs play an Online Strategy Game (Diplomacy) [Research Talk, NAACL 2025]

How to find and measure AI's weak spots [Research, NAACL 2025 Outstanding Paper]

Can AI write Questions that it can't Answer? [Research, NAACL 2025]

AIs can help you study if they can read your flashcards [Research, EMNLP 2024]

Computer Translation a few Words at a Time: Phrase-Based Models [Lecture]

The OG Machine Translation Model: IBM Model 1 [Lecture]

How AI can help you learn new vocabulary words [Research, EMNLP 2024]

How to turn trivia questions into state-of-the-art AI training data [Research, EMNLP 2024]

Discovering skills computers still lack when they try to answer questions. [Research, EMNLP 2024]

What’s the p(Doom)? And what’s my 5-year track record at AI Predictions?

How (Un-)Realistic is Dropout's "AI-generated" Video? [Rant]

Meta released a Diplomacy-playing LLM. How good is Cicero at talking to players? [Research Talk]

What Makes for a Bad Question? [Lecture]