INDEX

Explanations

attends to context-specific programming tokens from general code-related tokens

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

0:0.10

1:0.16

2:0.12

3:0.12

4:0.13

5:0.04

6:0.13

7:0.15

Negative Logits

 Salter

-0.30

 Micha

-0.29

ValueGenerated

-0.28

放題

-0.27

menade

-0.27

)(

-0.26

 היתה

-0.26

łę

-0.26

 obro

-0.25

 RestTemplate

-0.25

POSITIVE LOGITS

ProcessEvent

0.41

 незавершена

0.40

+#+#

0.38

NameInMap

0.37

 Houſe

0.36

 مشين

0.35

 Monfieur

0.35

OGND

0.35

{?>

0.35

 autorytatywna

0.35

Activations Density 0.003%

attends to context-specific programming tokens from general code-related tokens

No Comments

No Known Activations