INDEX

Explanations

attends to tokens indicating relationships or characteristics from tokens that specify conditions or actions

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

1:0.15

2:0.05

3:0.05

4:0.09

5:0.47

6:0.03

7:0.05

Negative Logits

retum

-0.38

 Audiodateien

-0.37

medriver

-0.37

StoryboardSegue

-0.37

endphp

-0.35

 ―――――

-0.34

 <<<<<<<<<<<<<<

-0.34

addGap

-0.34

 Autorizaciones

-0.34

amental

-0.34

POSITIVE LOGITS

Pierce

0.25

 esternos

0.25

 manera

0.24

 relever

0.24

 способ

0.24

Золо

0.23

 poveznice

0.23

ag

0.23

 racine

0.23

HideFlags

0.23

Activations Density 0.495%

attends to tokens indicating relationships or characteristics from tokens that specify conditions or actions

No Comments

No Known Activations