INDEX

Explanations

attends to numerical indicators marked with double asterisks from author names and brackets in the citations

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

0:0.06

1:0.07

2:0.07

3:0.09

4:0.09

5:0.05

6:0.42

7:0.13

Negative Logits

OGND

-0.53

TagMode

-0.45

endregion

-0.44

 Wikispecies

-0.43

PerformLayout

-0.42

Посилання

-0.39

abstractmethod

-0.39

Portail

-0.38

ⓘ

-0.37

 lenker

-0.37

POSITIVE LOGITS

tagHelper

0.33

StoryboardSegue

0.32

ayout

0.29

doria

0.29

 بخ

0.29

Попис

0.29

\%

0.28

kable

0.28

padek

0.28

ed

0.28

Activations Density 0.068%

attends to numerical indicators marked with double asterisks from author names and brackets in the citations

No Comments

No Known Activations