INDEX

Explanations

attends to the token "Read" from later tokens that suggest additional information or articles

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

2:0.05

3:0.02

4:0.01

5:0.01

6:0.04

7:0.03

8:0.01

9:0.01

10:0.01

11:0.01

12:0.04

13:0.62

14:0.05

15:0.01

Negative Logits

 Italij

-0.68

 Infór

-0.65

 zijne

-0.61

 desmotivaciones

-0.59

 Monfieur

-0.58

 miniaturka

-0.57

 Einfluß

-0.55

 čierna

-0.54

 Bewußt

-0.54

 Económica

-0.54

POSITIVE LOGITS

mand

0.41

bl

0.40

ob

0.39

ter

0.39

ins

0.39

lev

0.39

nar

0.39

ais

0.39

nal

0.38

Activations Density 0.109%

attends to the token "Read" from later tokens that suggest additional information or articles

No Comments

No Known Activations