INDEX

Explanations

attends to tokens associated with "LD" from tokens related to "ML"

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

0:0.16

1:0.22

2:0.12

3:0.07

4:0.10

5:0.11

6:0.04

7:0.14

Negative Logits

delig

-0.30

Бахар

-0.30

ownload

-0.30

führt

-0.29

OrderItem

-0.28

つき

-0.28

й

-0.28

OBL

-0.27

 Bade

-0.27

ode

-0.26

POSITIVE LOGITS

XmlAccessType

0.42

ListItemIcon

0.39

Datuak

0.39

 NgModule

0.35

 OnInit

0.35

oredCriteria

0.34

InSection

0.33

 snippetHide

0.33

 ddelweddau

0.33

 GenerationType

0.33

Activations Density 0.683%

attends to tokens associated with "LD" from tokens related to "ML"

No Comments

No Known Activations