INDEX

Explanations

attends to tokens associated with "inner" from tokens that denote a broader context or setting, which are earlier in the sequence

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.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

0:0.20

1:0.03

2:0.03

3:0.01

4:0.48

5:0.12

6:0.02

7:0.08

Negative Logits

 calendriers

-0.42

UnknownFields

-0.41

ItemBackground

-0.38

 يتيمه

-0.38

 nahilalakip

-0.37

Abit

-0.35

 fubject

-0.35

 gereken

-0.35

Geplaatst

-0.33

 hjel

-0.33

POSITIVE LOGITS

 nonatomic

0.44

corer

0.38

 مشين

0.36

ölker

0.36

athies

0.35

erai

0.35

луй

0.34

Искәрмәләр

0.34

 Superhosts

0.34

 rospy

0.34

Activations Density 0.018%

attends to tokens associated with "inner" from tokens that denote a broader context or setting, which are earlier in the sequence

No Comments

No Known Activations