INDEX

Explanations

attends to numbered tokens from formulation-related 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.09

1:0.17

2:0.12

3:0.12

4:0.14

5:0.15

6:0.08

7:0.10

Negative Logits

Personendaten

-0.47

isContained

-0.43

 AssemblyVersion

-0.42

expandindo

-0.41

principalTable

-0.41

 createSlice

-0.40

Vidite

-0.40

esgue

-0.40

AddTagHelper

-0.39

 NSCoder

-0.39

POSITIVE LOGITS

ély

0.26

...

0.24

 role

0.22

getRole

0.22

 will

0.20

Analyze

0.20

pédie

0.20

0.19

genommen

0.19

Activations Density 0.128%

attends to numbered tokens from formulation-related tokens

No Comments

No Known Activations