INDEX

Explanations

attends to specific tokens related to attributes or identifiers from subsequent tokens indicating their contexts or categories

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

0:0.11

1:0.10

2:0.40

3:0.07

4:0.03

5:0.02

6:0.05

7:0.18

Negative Logits

 herself

-0.52

afficheront

-0.40

 koja

-0.34

 princess

-0.33

 která

-0.33

herself

-0.33

 Kaur

-0.33

 Princess

-0.32

 ujednoznacz

-0.31

afterEach

-0.31

POSITIVE LOGITS

AddTagHelper

0.53

AsUp

0.41

MessageTagHelper

0.40

Higgins

0.39

 Référence

0.39

 său

0.38

 himself

0.38

InitVars

0.38

 CreateTagHelper

0.38

 dela

0.37

Activations Density 0.132%

attends to specific tokens related to attributes or identifiers from subsequent tokens indicating their contexts or categories

No Comments

No Known Activations