INDEX

Explanations

attends to acronyms or shortened forms from tokens that are elaborations or explanations of those abbreviations

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.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

0:0.81

1:0.00

2:0.01

3:0.01

4:0.09

5:0.02

6:0.01

7:0.02

Negative Logits

it

-0.24

our

-0.23

 Part

-0.22

for

-0.21

so

-0.20

We

-0.20

holdet

-0.20

DATABASE

-0.20

Inc

-0.20

fun

-0.19

POSITIVE LOGITS

 مشين

0.44

 AssemblyTitle

0.42

verwijspagina

0.41

saraba

0.40

AddTagHelper

0.37

 &___

0.35

 Италијани

0.35

UnknownFieldSet

0.35

PhysRev

0.34

 autorytatywna

0.33

Activations Density 6.715%

attends to acronyms or shortened forms from tokens that are elaborations or explanations of those abbreviations

No Comments

No Known Activations