INDEX

Explanations

No Explanations Found

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 recovered

-0.45

 recover

-0.45

σμο

-0.44

 recovering

-0.44

 contigo

-0.43

 freed

-0.43

 récupération

-0.41

 heavily

-0.41

 Begle

-0.41

 liberated

-0.41

POSITIVE LOGITS

 незавершена

0.81

parsedMessage

0.69

 queſta

0.62

 kasarigan

0.61

 pleaſure

0.59

0.57

ſelves

0.56

ſelf

0.56

MigrationBuilder

0.56

OCCURRED

0.55

Activations Density 0.000%

No Known Activations

This feature has no known activations.