INDEX

Explanations

attends to treatment-related tokens from specific coded or abbreviation 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

0:0.05

1:0.16

2:0.08

3:0.08

4:0.24

5:0.20

6:0.08

7:0.07

Negative Logits

BuilderFactory

-0.40

__*/

-0.37

 lenker

-0.37

 altrimenti

-0.35

NUMX

-0.34

:✨

-0.32

 originais

-0.31

InstrumentedTest

-0.31

 ivelany

-0.31

 morire

-0.31

POSITIVE LOGITS

querque

0.29

PhysRevLett

0.28

raborty

0.27

omycin

0.26

RTEX

0.26

 صوتيه

0.25

ocities

0.25

 colspan

0.25

vied

0.24

ftrag

0.24

Activations Density 0.832%

attends to treatment-related tokens from specific coded or abbreviation tokens

No Comments

No Known Activations