INDEX

Explanations

attends to numeric tokens from broader summary or topic-related 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

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.09

1:0.12

2:0.13

3:0.09

4:0.08

5:0.05

6:0.10

7:0.30

Negative Logits

WebElementEntity

-0.33

ביוגרפיה

-0.32

Rüyada

-0.32

AndEndTag

-0.32

 informée

-0.31

 OMITBAD

-0.30

Gön

-0.29

 Bourgoin

-0.29

 برانيه

-0.29

Calais

-0.27

POSITIVE LOGITS

ConstraintMaker

0.37

aneous

0.26

matician

0.26

 ویکی‌پدیای

0.26

ρέπει

0.24

installa

0.23

vellous

0.23

 adhered

0.23

izacao

0.23

jectory

0.23

Activations Density 0.022%

attends to numeric tokens from broader summary or topic-related tokens

No Comments

No Known Activations