INDEX

Explanations

attends to instances of "rhs" from instances of any token

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

0:0.07

1:0.09

2:0.07

3:0.28

4:0.10

5:0.03

6:0.20

7:0.11

Negative Logits

TI

-0.30

Bern

-0.30

一刻

-0.29

 meines

-0.28

Ti

-0.27

 Tout

-0.27

TTI

-0.27

 dene

-0.26

rav

-0.26

 Bern

-0.26

POSITIVE LOGITS

 propOrder

0.38

ագրություններ

0.37

 autorytatywna

0.37

!*\

0.36

IonicModule

0.35

autoreleasepool

0.35

HostException

0.34

 onAnimation

0.33

 vVar

0.32

Espèce

0.32

Activations Density 0.160%

attends to instances of "rhs" from instances of any token

No Comments

No Known Activations