INDEX

Explanations

attends to the last token in a sequence from various preceding 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

0:0.11

1:0.12

2:0.10

3:0.13

4:0.15

5:0.07

6:0.16

7:0.12

Negative Logits

 wireType

-0.42

 useAppContext

-0.42

 مرئيه

-0.39

tonode

-0.36

RetentionPolicy

-0.36

Abitanti

-0.34

 Forumite

-0.34

<?

-0.33

 parseFrom

-0.33

HasAnnotation

-0.32

POSITIVE LOGITS

<eos>

0.44

nog

0.33

</em>

0.33

Viitteet

0.31

ggior

0.31

wapV

0.30

ar

0.28

il

0.28

errHandler

0.27

<em>

0.27

Activations Density 0.099%

attends to the last token in a sequence from various preceding tokens

No Comments

No Known Activations