INDEX

Explanations

attends to familial or relational terms from corresponding 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

0:0.10

1:0.10

2:0.52

3:0.06

4:0.05

5:0.02

6:0.04

7:0.07

Negative Logits

ddots

-0.30

perror

-0.28

BPF

-0.27

otides

-0.26

/=

-0.25

harusnya

-0.25

pfung

-0.25

tze

-0.25

Prin

-0.25

↩

-0.25

POSITIVE LOGITS

 jsPsych

0.33

ที่ยว

0.32

Pozri

0.31

 ImportError

0.31

 estekak

0.30

Билгалдахарш

0.30

essoal

0.29

 Seeder

0.29

 εί

0.29

 snippetHide

0.29

Activations Density 0.330%

attends to familial or relational terms from corresponding tokens

No Comments

No Known Activations