INDEX

Explanations

attends to tokens that express gratitude or acknowledgment from later 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

0:0.09

1:0.12

2:0.35

3:0.11

4:0.07

5:0.04

6:0.07

7:0.12

Negative Logits

 protoimpl

-0.39

SharedCtor

-0.36

 الاطلاع

-0.36

LikeLike

-0.36

AddHtmlAttribute

-0.35

 sumpay

-0.34

LikeLiked

-0.33

SharedDtor

-0.32



-0.31

chtenstein

-0.30

POSITIVE LOGITS

IsContent

0.27

 disambiguazione

0.27

colgroup

0.27

 Dies

0.25

SAX

0.25

gew

0.25

Enqueue

0.24

Pop

0.24

rzost

0.23

поло

0.22

Activations Density 0.160%

attends to tokens that express gratitude or acknowledgment from later tokens

No Comments

No Known Activations