INDEX

Explanations

attends to ambiguous phrases or tokens from related syntactical markers or programming constructs

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

0:0.48

1:0.20

2:0.08

3:0.05

4:0.04

5:0.03

6:0.03

7:0.04

Negative Logits

findpost

-0.30

 hiver

-0.28

цездатний

-0.27

 виправивши

-0.27

tvguidetime

-0.26

 compro

-0.25

 старости

-0.25

Salta

-0.24

msgTypes

-0.24

 MotionEvent

-0.24

POSITIVE LOGITS

rrggbb

0.33

 seguida

0.33

 hâte

0.33

ucson

0.30

Kru

0.30

łosi

0.30

 Watanabe

0.30

 ioutil

0.30

marginHorizontal

0.29

 Cous

0.29

Activations Density 0.706%

attends to ambiguous phrases or tokens from related syntactical markers or programming constructs

No Comments

No Known Activations