Сохраненные ссылки

Рабочая свалка полезных ссылок, справочников, инструментов и заметок.

Harness Engineering for Self-Improvement

Lilian Weng on agent harnesses, context and memory management, workflow design, and self-improvement.

agents llm research

Interactive exercises on transformer internals, from attention and RoPE to KV caches and compute costs.

transformers llm interactive

A guide from the Kuleshov Group to building diffusion language models.

diffusion llm training

Fine-tuning a small language model for structured outputs with GRPO, TRL, and the IFStruct dataset.

grpo llm training