INDEX

Explanations

attends from numeric tokens in sequences to their immediate preceding numeric 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

0:0.07

1:0.08

2:0.13

3:0.41

4:0.08

5:0.06

6:0.07

7:0.07

Negative Logits

featureID

-0.50

 مشين

-0.48

expandindo

-0.46

tonsoft

-0.46

 ModelExpression

-0.44

 Wikimedijinoj

-0.43

 JAXBElement

-0.42

contentLoaded

-0.40

basicConfig

-0.39

 تضيفلها

-0.39

POSITIVE LOGITS

 again

0.34

Again

0.33

 Again

0.32

again

0.31

 nuovamente

0.30

vidla

0.29

 AGAIN

0.28

AGAIN

0.27

 intents

0.26

 pomy

0.26

Activations Density 0.929%

attends from numeric tokens in sequences to their immediate preceding numeric tokens

No Comments

No Known Activations