working · Nick Meyer

Replicated IFT, developed an improved objective function, and fitted jacobian lenses on Llama 3.2 and Qwen3.

Finetuning Llama 3.2 1B with SFT and DPO + LoRA for improved harmlessness.

My work through the ARENA curriculum in interpretability and alignment as part of CBAI's CAMBRIA program.

Structural misalignment and the limit of technical approaches.

A comparison of PCA and linear autoencoders on MNIST.

Artificial intelligence, alignment, and building a future worth wanting.

Mechanistic interpretability as the path forward towards achieving a duty of justification.

A misdiagnosis of power and Foucault's challenge to domination theory.