INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

gradle

-0.46

SqlCommand

-0.41

 gezet

-0.40

rigen

-0.40

 hautes

-0.39

věr

-0.39

masalahan

-0.38

expandindo

-0.37

 Persson

-0.37

 courrier

-0.37

POSITIVE LOGITS

 esboço

5.44

 незавершена

1.27

 sketch

0.93

 esbo

0.90

sketch

0.86

 sketches

0.81

 '\\;'

0.77

httphttps

0.76

 sketching

0.76

Sketch

0.76

Activations Density 0.000%

No Known Activations

This feature has no known activations.