INDEX

Explanations

attends to utility-related tokens from preceding import tokens

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 10-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_10/width_16k/average_l0_70

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.10.attn.hook_z

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

Head Attr Weights

0:0.13

1:0.15

2:0.28

3:0.14

4:0.10

5:0.03

6:0.04

7:0.08

Negative Logits

 Bourgoin

-0.30

Wir

-0.25

 Basta

-0.24

 Proceed

-0.23

าศ

-0.23

far

-0.23

SOS

-0.23

setVerticalGroup

-0.22

Ris

-0.22

 proceed

-0.22

POSITIVE LOGITS

IsMutable

0.41

IntoConstraints

0.40

KURZBESCHREIBUNG

0.39

 estimés

0.33

 Glied

0.32

 ThemeData

0.32

HasAnnotation

0.32

AnimationsModule

0.32

 snippetHide

0.31

ACIONAL

0.30

Activations Density 0.014%

attends to utility-related tokens from preceding import tokens

No Comments

No Known Activations