INDEX

Explanations

attends to information-related tokens from adjacent location-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

1:0.11

2:0.10

3:0.11

4:0.10

5:0.05

6:0.18

7:0.23

Negative Logits

stok

-0.27

by

-0.26

...

-0.25

gina

-0.25

ुष

-0.24

Ger

-0.24

tak

-0.24

ViewById

-0.24

sig

-0.24

dec

-0.23

POSITIVE LOGITS

 CreateTagHelper

0.58

DockStyle

0.51

RetentionPolicy

0.50

rungsseite

0.48

nodoc

0.47

 مرئيه

0.45

ientôt

0.44

úgó

0.44

StructEnd

0.44

 ModelExpression

0.44

Activations Density 0.318%

attends to information-related tokens from adjacent location-related tokens

No Comments

No Known Activations