Beyond One Output: Visualizing and Comparing Distributions of Language Model Generations

Build a tool to visualize LLM output distributions beyond the single top-k token. This helps prompt engineers debug sensitivity and model confidence.

Suggested repo: DistriView

"Stop debugging with single outputs; visualize the full probability distribution of your LLM."

Estimated effort: 60h