enbao.me
posts from enbao.me
Latest posts
an ocean, a reminder
the ocean and the point of it all
postcard from china
reflections from a trip to china
new stars in my sky
reflections on childhood and being at the beginning
home for the holidays
the little things
for granted
a healthy man wants a thousand things, and a sick man just wants one
the bay
reflections after a week in the bay
November Paper Reading
detecting adversarial fine-tuning, litmus values, impossiblebench
like dreams
see as much as you can, as early as you can
October Paper Reading
more work on agents, bridging the non-verifiable gap, sft vs rl, objective function variants, scheming, lora again
hear the wind sing
on being shaped by the writers you love
September Paper Reading
prompt evolution, world models, moe efficiency, synthetic data pipeline
Late August Paper Reading
a writing model, rubrics, data filtering, hallucination detection, entropy-based confidence
Mid August Paper Reading
glm 4.5, persona vectors, and system cards
Early August Paper Reading
kimi k2, reasoning length, system prompts, dual-token rlvr
Optimizers
looking into and testing sgd, momentum, rmsprop, adam, adamw, and muon
Late July Paper Reading
toxic data, alignment faking, dynamic chunking, entropy-based exploration, general reward models, magistral
Mid July Paper Reading
seed1.5, generalist reward models, spiral, natural thoughts, 2-simplical attention
Early July Paper Reading
autoregressive u-nets, diffusion llms, test-time compute, rlvr, and mechanistic interpretability
Inside ThunderKittens' Python Bindings
how ThunderKittens wraps CUDA kernels for Python