INDEX

Explanations

modelnatural language

This neuron fires on product and model identifier tokens—especially capitalized item descriptors and numbers in e-commerce style titles.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 tetragonal

0.71

 bract

0.67

 grafts

0.66

 ఎక్కువ

0.65

 getN

0.65

 uncoated

0.65

 dichas

0.64

 manipulations

0.64

 sprzę

0.64

 permut

0.63

POSITIVE LOGITS

Когда

0.69

Cuando

0.68

Humans

0.68

Wonderful

0.67

Investing

0.67

When

0.66

Consumers

0.66

Take

0.66

Imagine

0.66

Containers

0.66

Activations Density 0.011%