INDEX

Explanations

attends to mathematical operators or symbols from numerical 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

2:0.09

3:0.10

4:0.09

5:0.06

6:0.21

7:0.23

Negative Logits

any

-0.26

emd

-0.26

fahrens

-0.25

 pena

-0.25

arda

-0.25

 того

-0.23

 đốc

-0.23

cula

-0.23

št

-0.23

nov

-0.23

POSITIVE LOGITS

WriteBarrier

0.48

LookAnd

0.44

 Normdatei

0.40

AccessorTable

0.36

 myſelf

0.36

iastes

0.35

 kasarigan

0.35

 itſelf

0.35

tagHelperRunner

0.34

 defaultstate

0.34

Activations Density 0.147%

attends to mathematical operators or symbols from numerical tokens

No Comments

No Known Activations