INDEX

Explanations

attends to the token "type" from numerical tokens or labels that specify attributes

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

0:0.49

1:0.20

2:0.06

3:0.05

4:0.04

5:0.03

6:0.03

7:0.05

Negative Logits

Xaml

-0.22

 they

-0.20

 served

-0.20

 restos

-0.18

ANTY

-0.18

引き

-0.18

łość

-0.18

kiwa

-0.18

规模

-0.17

iale

-0.17

POSITIVE LOGITS

verwijspagina

0.38

IsMutable

0.35

 kasarigan

0.32

endphp

0.30

EDEFAULT

0.29

存于互联网档案馆

0.29

 تضيفلها

0.28

󠁿

0.28

 مرئيه

0.26

}});

0.26

Activations Density 0.865%

attends to the token "type" from numerical tokens or labels that specify attributes

No Comments

No Known Activations