Egor PakhomovRLHF based on humans just normally talking to LLMRLHF always rely on humans explicitly evaluating and training the model. Normal LLM users just talking to the model contains signal as…Apr 12, 2023Apr 12, 2023
Egor PakhomovBeyond Pre-Training and Prompt Engineering: The Value of LLM Fine-TuningAdvancements in LLM focus on pre-training and prompt engineering, while fine-tunning is being under-invested in.Apr 11, 20231Apr 11, 20231
Egor PakhomovLimitations of Prompt Engineering in Context-Heavy ScenariosUsing the GPT prompt in its simplest form involves posing a question to the model, such as “How do I make cookies?” or “How do you write…Mar 24, 2023Mar 24, 2023
Egor PakhomovОтзыв об четырех-дневном водяном голоданииЯ только что закончил 4.5 суточное(108 часов) голодание на воде. Мне очень понравился опыт и я хочу им поделиться.Jun 23, 20191Jun 23, 20191
Egor PakhomovColorado review from Tahoe regular skierSking in Colorado is great. However there are several things you need to take into account before planning your trip…Feb 10, 20191Feb 10, 20191
Egor PakhomovSpark Datasets are not as type-safe as you thinkWhen Dataset API was introduced in Spark 2.0 the value proposition was the ability to use functionality of Dataframes in type-safe manner…Jan 5, 20191Jan 5, 20191