INDEX

Explanations

attends to unlimited-related tokens from tokens that are associated with specific features or applications

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

0:0.08

1:0.11

2:0.08

3:0.11

4:0.11

5:0.07

6:0.21

7:0.18

Negative Logits

resizingMask

-0.31

 Meksiku

-0.27

 ويكيپيديا

-0.26

 Ludo

-0.24

grao

-0.24

—

-0.24

CFS

-0.24

 ostavi

-0.23

 remotely

-0.23

odeon

-0.23

POSITIVE LOGITS

 berta

0.30

 CreateTagHelper

0.29

vă

0.28

 économies

0.28

themselves

0.28

 kasar

0.28

domésticos

0.28

 pulseras

0.26

sī

0.26

 AssemblyCompany

0.25

Activations Density 0.071%

attends to unlimited-related tokens from tokens that are associated with specific features or applications

No Comments

No Known Activations