INDEX

Explanations

attends to numbers from tokens containing the same index

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.08

1:0.14

2:0.08

3:0.03

4:0.06

5:0.47

6:0.03

7:0.06

Negative Logits

ly

-0.25

火

-0.23

 signs

-0.23

 signes

-0.23

vač

-0.23

Ch

-0.21

 show

-0.21

Ch

-0.20

 conden

-0.20

 sign

-0.20

POSITIVE LOGITS

 doInBackground

0.45

GEBURTSDATUM

0.44

 Савезне

0.44

richTextPanel

0.41

 AssemblyTitle

0.40

-------

0.40

تقاوى

0.40

RectangleBorder

0.40

 CreateTagHelper

0.38

 disambiguazione

0.38

Activations Density 0.095%

attends to numbers from tokens containing the same index

No Comments

No Known Activations