INDEX

Explanations

attends to tokens that are standalone or compound structures marked in uniform syntax from those that appear as delimiter characters or structures in code snippets

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.06

1:0.10

2:0.09

3:0.10

4:0.08

5:0.06

6:0.32

7:0.16

Negative Logits

su

-0.24

endregion

-0.24

周

-0.23

<u>

-0.23

 thế

-0.23

گزاری

-0.22

-0.21

bes

-0.21

hadoop

-0.21

pro

-0.21

POSITIVE LOGITS

 SwitchCompat

0.38

WriteAttribute

0.37

وصلات

0.37

rungsseite

0.37

 pinulongan

0.35

MOP

0.35

 Waray

0.35

ьаж

0.35

BEA

0.35

 ujednoznacz

0.34

Activations Density 0.024%

attends to tokens that are standalone or compound structures marked in uniform syntax from those that appear as delimiter characters or structures in code snippets

No Comments

No Known Activations