Experiment 28
I tested four bandit algorithms on news recommendation to figure out if contextual learning could teach LLMs to adapt their prompting style to individual use...
Recent Posts
Experiment 27
This post presents an inventory tracking simulation that demonstrates how Kalman filters optimally estimate total inventory across partially-observable shelv...
Experiment 26
This post analyzes urban mobility patterns to identify critical mobility hubs, network resilience, and neighborhood structures within cities by applying cent...
Experiment 25
This post demonstrates that reinforcement learning agents can be trained with sparse reward signals as effectively as carefully tuned dense rewards.
Experiment 24
The blog post demonstrates that few-shot LLMs can match 96% of fine-tuned model accuracy on scene graph extraction with 103x faster inference.