INDEX

Explanations

attends to tokens related to "leading" from any subsequent token that is a noun or a noun phrase

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-att/layer_0/width_16k/average_l0_61

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.attn.hook_z

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.01

1:0.01

2:0.53

3:0.02

4:0.00

5:0.00

6:0.09

7:0.04

8:0.00

9:0.01

10:0.00

11:0.00

12:0.01

13:0.11

14:0.07

15:0.00

Negative Logits

 calendriers

-0.26

 chłop

-0.21

 Wikiseite

-0.20

 zboží

-0.20

 varandra

-0.19

 avoient

-0.19

 habet

-0.18

 miniaturka

-0.18

DEPORTES

-0.18

carteles

-0.18

POSITIVE LOGITS

Personensuche

0.34

memset

0.32

❹

0.30

Nazionale

0.30

SPJ

0.29

ArrowToggle

0.29

 Wass

0.28

Synopsis

0.28

 Anywhere

0.28

Ang

0.28

Activations Density 0.101%

attends to tokens related to "leading" from any subsequent token that is a noun or a noun phrase

No Comments

No Known Activations