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

re

-0.32

over

-0.32

<eos>

-0.31

//$

-0.29

//

-0.28

 seems

-0.28

ade

-0.28

ual

-0.27

so

-0.27

 made

-0.27

POSITIVE LOGITS

Controllers

2.02

 Controllers

1.45

 controllers

1.24

 Controller

1.16

controllers

1.15

 controller

1.06

 ControllerBase

1.03

Controller

1.02

controller

1.01

 UserController

0.94

Activations Density 0.000%

No Known Activations

This feature has no known activations.