INDEX

Explanations

attends to category-related tokens from other category-indicative tokens

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

0:0.09

1:0.11

2:0.11

3:0.23

4:0.09

5:0.05

6:0.14

7:0.14

Negative Logits

Попис

-0.39

niająca

-0.31

Autoritní

-0.30

transQ

-0.28

Predecesor

-0.27

ENDIAN

-0.27

IntoConstraints

-0.27

LookAnd

-0.27

tský

-0.27

contentLoaded

-0.26

POSITIVE LOGITS

fraid

0.35

umente

0.33

 Pity

0.32

abetes

0.30

 armées

0.30

 محفوظة

0.30

 officinalis

0.30

 Paro

0.29

cinfo

0.29

asants

0.28

Activations Density 0.000%

attends to category-related tokens from other category-indicative tokens

No Comments

No Known Activations