INDEX

Explanations

providing specific examples

This neuron detects copular constructions where a subject (often “She” or “He”) is followed by one or more descriptive adjectives or praise-type attributes (e.g. “She is beautiful,” “She is calm,” “She is amazing”).

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

 Valves

0.23

 Coalition

0.23

 Williams

0.22

 Compound

0.22

 Preset

0.22

 Canal

0.22

 Peak

0.22

 RISK

0.21

 inny

0.21

 Pulse

0.21

POSITIVE LOGITS

↵↵↵↵↵↵↵↵↵↵

0.28

𒉰

0.27

依旧

0.27

áh

0.27

árez

0.26

Ļ

0.26

 परंपरागत

0.25

𒐪

0.25

ছিলো

0.25

 đươ

0.25

Activations Density 1.583%