INDEX

Explanations

attends to position-related tokens from the subsequent numerical rank 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

1:0.12

2:0.12

3:0.12

4:0.10

5:0.07

6:0.14

7:0.17

Negative Logits

=\""

-0.35

 ―――――

-0.33

\""

-0.33

."</

-0.33

脚注の使い方

-0.33

]='\

-0.32

pshots

-0.31

ביוגרפיה

-0.31

PerformLayout

-0.31

 ――――

-0.31

POSITIVE LOGITS

UnitTesting

0.26

NameInMap

0.25

 useAppContext

0.24

stmt

0.24

featureID

0.23

ness

0.23

 того

0.22

mpä

0.22

ya

0.22

 confirmer

0.22

Activations Density 0.010%

attends to position-related tokens from the subsequent numerical rank tokens

No Comments

No Known Activations