PinnedPublished inAI AdvancesYou Don’t Always Need An MCP ClientUsing custom HTTP clients to interact with MCP servers5d ago15d ago1
PinnedPublished inAI AdvancesThe Biggest Enemy of Generative ModelingClever ways in which modern models have dodged the evils of parameter estimationApr 173Apr 173
PinnedPublished inAI AdvancesIs your LLM “playing along” with your guardrails?Understanding Alignment Faking in Large Language Models and Ways to Avoid ItMar 54Mar 54
PinnedPublished inAI AdvancesHow did we get to vLLM, and what was its genius?The evolution of the model serving landscapeFeb 151Feb 151
PinnedPublished inAI AdvancesTransformer²: The Death of LoRA?Self-Adaptive LLMs with Singular Value Fine-tuningJan 192Jan 192
Stop treating your LLM like a black box, you MORON!Break open your models with Mechanistic Interpretability.Jan 17Jan 17
Chain-of-Thought — All the Language Models are Just Extremely Intelligent ToddlersAug 7, 2024Aug 7, 2024
