INDEX

Explanations

attends to various token patterns that may denote the attributes or actions of players or entities in a competitive context from related succeeding 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.22

1:0.21

2:0.11

3:0.08

4:0.13

5:0.06

6:0.06

7:0.11

Negative Logits

stå

-0.35

Micah

-0.34

 Gale

-0.34

Vay

-0.34

 Đường

-0.33

 đường

-0.33

 вік

-0.33

Gale

-0.32

 fixé

-0.32

fæ

-0.32

POSITIVE LOGITS

 AssemblyTitle

0.45

########.

0.43

 ***!

0.41

 AssemblyCulture

0.40

存于互联网档案馆

0.40

 okuyayım

0.38

 Hamp

0.37

 digress

0.36

mipmap

0.36

ControllerAdvice

0.35

Activations Density 0.301%

attends to various token patterns that may denote the attributes or actions of players or entities in a competitive context from related succeeding tokens

No Comments

No Known Activations