INDEX

Explanations

attends to tokens related to programming constructs from tokens related to web architecture

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

0:0.11

1:0.19

2:0.22

3:0.10

4:0.14

5:0.05

6:0.07

7:0.08

Negative Logits

sizeCache

-0.35

 initComponents

-0.34

InvalidProtocol

-0.34

الحياه

-0.33

ParseTree

-0.32

prüfung

-0.32

]='\

-0.32

 labours

-0.31

strecke

-0.31

 StatelessWidget

-0.31

POSITIVE LOGITS

án

0.27

ZoneId

0.24

zenta

0.24

 trin

0.24

dez

0.23

faits

0.23

úgó

0.23

 vang

0.23

cinogenicity

0.23

 }}$}

0.23

Activations Density 0.001%

attends to tokens related to programming constructs from tokens related to web architecture

No Comments

No Known Activations