INDEX

Explanations

attends to tokens indicating numerical values or classifications from related descriptive 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

0:0.23

1:0.26

2:0.13

3:0.09

4:0.08

5:0.02

6:0.05

7:0.10

Negative Logits

LookAnd

-0.28

AddHtmlAttribute

-0.28

 beginnetje

-0.26

хьтан

-0.26

delwed

-0.25

keyColumn

-0.24

 AssemblyCulture

-0.24

igt

-0.23

 ProtoMessage

-0.23

новниш

-0.22

POSITIVE LOGITS

 للمعارف

0.48

 myſelf

0.39

 بيها

0.38

 RIPRODUZIONE

0.35

 pleaſure

0.34

}))

0.34

 Majefty

0.33

 betweenstory

0.33

 itſelf

0.33

 ARXIV

0.33

Activations Density 12.227%

attends to tokens indicating numerical values or classifications from related descriptive tokens

No Comments

No Known Activations