INDEX

Explanations

attends to certain pairs of tokens that signify emphasis or connections, particularly focusing on colons and conjunctions that link various elements

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

1:0.11

2:0.11

3:0.15

4:0.08

5:0.07

6:0.14

7:0.19

Negative Logits

 generali

-0.31

contentLoaded

-0.29

Šaltiniai

-0.29

ená

-0.28

 mergeFrom

-0.28

MigrationBuilder

-0.27

and

-0.27

webElementXpaths

-0.26

ThroughAttribute

-0.26

LookAnd

-0.26

POSITIVE LOGITS

存于互联网档案馆

0.45

nastics

0.44

ArgsConstructor

0.42

')[

0.40

SOF

0.40

]='\

0.40

 diamonds

0.40

capulco

0.39

testens

0.39

guenos

0.38

Activations Density 0.000%

attends to certain pairs of tokens that signify emphasis or connections, particularly focusing on colons and conjunctions that link various elements

No Comments

No Known Activations