INDEX

Explanations

attends to the general token "works" from tokens that relate to specific contexts in literature

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

1:0.12

2:0.12

3:0.13

4:0.18

5:0.10

6:0.13

7:0.12

Negative Logits

 ویکی‌پدیای

-0.28

lode

-0.22

 chêne

-0.22

startswith

-0.21

ym

-0.21

啱

-0.21

,"

-0.21

PathVariable

-0.21

",&

-0.21

Password

-0.21

POSITIVE LOGITS

 resourceCulture

0.42

 فريبيس

0.41

Identyfik

0.37

DataAnnotations

0.37

 ComVisible

0.36

 myſelf

0.35

InitVars

0.35

adpleegd

0.34

 Мексичка

0.33

 anthrop

0.33

Activations Density 0.739%

attends to the general token "works" from tokens that relate to specific contexts in literature

No Comments

No Known Activations