INDEX

Explanations

attends to scientific and technical terms or phrases from corresponding markers or references related to those terms

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

0:0.33

1:0.29

2:0.10

3:0.06

4:0.05

5:0.01

6:0.04

7:0.07

Negative Logits

^(@)

-0.42

ERÍA

-0.41

."</

-0.41

ohner

-0.39

)++;

-0.37

 minY

-0.37

šanai

-0.36

YNC

-0.36

 onCancelled

-0.36

ENEFITS

-0.36

POSITIVE LOGITS

DoubleQuotes

0.33

encodeWith

0.33

 typelib

0.32

ScopeManager

0.31

lccn

0.29

Personensuche

0.29

AnchorTagHelper

0.29

 ellas

0.28

τας

0.28

LookAnd

0.27

Activations Density 6.755%

attends to scientific and technical terms or phrases from corresponding markers or references related to those terms

No Comments

No Known Activations