INDEX

Explanations

attends to object tokens that are formatted in specific documentations rules while being associated with state 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

1:0.14

2:0.14

3:0.09

4:0.08

5:0.07

6:0.16

7:0.17

Negative Logits

帖最后由

-0.34

saraba

-0.34

 Winaray

-0.34

photobucket

-0.32

CppMethod

-0.30

#+#

-0.30

AndEndTag

-0.29

gios

-0.29

setViewportView

-0.27

multer

-0.27

POSITIVE LOGITS

InjectAttribute

0.34

gages

0.28

 aveug

0.27

 مشين

0.27

httphttps

0.26

HasIndex

0.25

 Chwiliwch

0.25

 مرئيه

0.25

expandindo

0.24

вах

0.24

Activations Density 0.072%

attends to object tokens that are formatted in specific documentations rules while being associated with state tokens

No Comments

No Known Activations