INDEX

Explanations

attends to action-related tokens from corresponding action-type 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

1:0.10

2:0.52

3:0.08

4:0.05

5:0.02

6:0.03

7:0.07

Negative Logits

EndProject

-0.42

]='\

-0.34

 kasarigan

-0.32

 Emin

-0.32

CppMethod

-0.32

Portail

-0.31

 stoppage

-0.31

WAUKEE

-0.30

 }}"></

-0.30

 consultato

-0.30

POSITIVE LOGITS

hasMoreElements

0.37

mapStateToProps

0.31

Demografía

0.31

 متعلقه

0.30

 createSlice

0.30

 wireType

0.30

se

0.29

esez

0.29

<?

0.28

ClassNotFound

0.27

Activations Density 0.391%

attends to action-related tokens from corresponding action-type tokens

No Comments

No Known Activations