INDEX

Explanations

attends to numeric values in mathematical expressions from preceding numeric values

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

0:0.07

1:0.09

2:0.08

3:0.11

4:0.10

5:0.08

6:0.30

7:0.14

Negative Logits

__':

-0.47

ViewFeatures

-0.45

omenclature

-0.43

 ویکی‌پدیا

-0.40

GraphicsUnit

-0.40

 المعيارى

-0.40

 OMITBAD

-0.40

TagHelper

-0.40

EDEFAULT

-0.39

Rhestr

-0.39

POSITIVE LOGITS

lds

0.32

תה

0.31

 olay

0.30

ledning

0.30

によ

0.29

kæ

0.29

 manera

0.28

raza

0.28

 grise

0.28

zins

0.28

Activations Density 0.037%

attends to numeric values in mathematical expressions from preceding numeric values

No Comments

No Known Activations