INDEX

Explanations

attends to the token "values" from token-related options

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

0:0.38

1:0.22

2:0.10

3:0.06

4:0.06

5:0.03

6:0.04

7:0.07

Negative Logits

kháu

-0.24

Neither

-0.23

 neither

-0.21

 doña

-0.20

Примітки

-0.20

 transcur

-0.20

ایج

-0.20

widgets

-0.20

sizeCache

-0.19

lighed

-0.19

POSITIVE LOGITS

Diweddarwch

0.28

modelBuilder

0.26

httphttps

0.26

 EconPapers

0.25

 Verk

0.24

 createSlice

0.24

)}=\

0.24

__':

0.24

RTEE

0.23

'}),

0.23

Activations Density 0.477%

attends to the token "values" from token-related options

No Comments

No Known Activations