INDEX

Explanations

attends to specific emotional or conceptual tokens from broader contextual 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

1:0.09

2:0.11

3:0.15

4:0.18

5:0.10

6:0.15

7:0.09

Negative Logits

 extra

-0.28

-0.26

AspNetCore

-0.26

,"

-0.25

ur

-0.25

-0.24

elf

-0.24

try

-0.24

if

-0.24

POSITIVE LOGITS

IntoConstraints

0.48

exitRule

0.47

 سكانية

0.46

 tartalomajánló

0.45

setVerticalGroup

0.45

ImageContext

0.43

ніципалі

0.43

 виправивши

0.43

djangoproject

0.41

 defaultstate

0.40

Activations Density 0.177%

attends to specific emotional or conceptual tokens from broader contextual tokens

No Comments

No Known Activations