INDEX

Explanations

attends to tokens indicating size or quantity from tokens indicating dimensions or structures

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

1:0.10

2:0.51

3:0.04

4:0.04

5:0.06

6:0.03

7:0.09

Negative Logits

躇

-0.44

errHandler

-0.39

最快更新

-0.38

 saites

-0.36

Heim

-0.34

 تضيفلها

-0.33

 Heim

-0.32

 دیکھیے

-0.32

subsection

-0.32

 μυ

-0.32

POSITIVE LOGITS

WriteAttribute

0.30

Примітки

0.28

Chham

0.27

clusal

0.27

 Pussy

0.27

vosť

0.27

 convertView

0.25

FormatException

0.25

 galloping

0.25

 melons

0.24

Activations Density 0.366%

attends to tokens indicating size or quantity from tokens indicating dimensions or structures

No Comments

No Known Activations