INDEX

Explanations

attends to specific geographical locations or entities from broader categories or titles related to them

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

0:0.12

1:0.23

2:0.22

3:0.06

4:0.10

5:0.04

6:0.07

7:0.13

Negative Logits

świad

-0.24

 rangka

-0.23

 call

-0.23

 Henderson

-0.22

Contents

-0.22

powiedz

-0.22

ampung

-0.21

 native

-0.21

 tiên

-0.21

rechter

-0.20

POSITIVE LOGITS

addCriterion

0.44

FormTagHelper

0.41

 ویکی‌پدی

0.41

UrlResolution

0.40

 Numerade

0.39

featureID

0.38

AddTagHelper

0.38

Personendaten

0.38

ulemon

0.36

 الرياضيه

0.35

Activations Density 0.589%

attends to specific geographical locations or entities from broader categories or titles related to them

No Comments

No Known Activations