INDEX

Explanations

attends to tokens with an underscore followed by any character from tokens marked with square brackets

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

0:0.23

1:0.05

2:0.04

3:0.06

4:0.35

5:0.10

6:0.07

7:0.06

Negative Logits

 Chriftian

-0.54

 sauvages

-0.50

 Monfieur

-0.50

 greateſt

-0.50

 Cæsar

-0.49

 fubject

-0.49

 Shakspeare

-0.49

 ſame

-0.48

ArgsConstructor

-0.48

 poffe

-0.47

POSITIVE LOGITS

 then

0.35

an

0.35

far

0.33

0.32

qu

0.32

 more

0.31

in

0.31

to

0.31

Activations Density 0.025%

attends to tokens with an underscore followed by any character from tokens marked with square brackets

No Comments

No Known Activations