INDEX

Explanations

attends to tokens that serve as markers or separators from the adjacent empty 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

0:0.10

1:0.19

2:0.10

3:0.07

4:0.24

5:0.14

6:0.06

7:0.07

Negative Logits

 resourceCulture

-0.47

HasForeignKey

-0.35

EndInit

-0.35

StoryboardSegue

-0.32

expandindo

-0.32

+#+

-0.31

 ExecuteAsync

-0.31

 createState

-0.30

AddTagHelper

-0.30

 transférez

-0.29

POSITIVE LOGITS

ronym

0.28

 köpa

0.28

 damit

0.25

 torr

0.24

 Winaray

0.24

itim

0.24

mybatisplus

0.24

DNF

0.23

 Freund

0.23

koop

0.23

Activations Density 0.001%

attends to tokens that serve as markers or separators from the adjacent empty tokens

No Comments

No Known Activations