INDEX

Explanations

attends to numerical tokens from corresponding ordered sequence indicators

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

0:0.05

1:0.07

2:0.05

3:0.09

4:0.07

5:0.03

6:0.51

7:0.10

Negative Logits

-0.37

-0.34

 diatas

-0.33

 guys

-0.32

raw

-0.32

️

-0.30

 FUCKING

-0.30

-0.29

//

-0.29

POSITIVE LOGITS

verwijspagina

0.61

 Meksiku

0.52

Vidite

0.50

—

0.49

Portale

0.47

 AssemblyCulture

0.46

 ويكيپيديا

0.45

contentLoaded

0.45

 betweenstory

0.45

ItemBackground

0.45

Activations Density 0.113%

attends to numerical tokens from corresponding ordered sequence indicators

No Comments

No Known Activations