INDEX

Explanations

attends to negated or contrasting phrases from their preceding positive statements

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

0:0.04

1:0.05

2:0.03

3:0.02

4:0.23

5:0.53

6:0.03

7:0.02

Negative Logits

and

-0.73

 because

-0.44

but

-0.43

 that

-0.38

<eos>

-0.38

 while

-0.37

or

-0.36

 however

-0.34

 since

-0.33

POSITIVE LOGITS

GEBURTSDATUM

0.52

astéroïdes

0.48

Vidite

0.48

 فريبيس

0.47

 poveznice

0.46

 مرئيه

0.45

Rhestr

0.45

Personensuche

0.45

WriteBarrier

0.44

̍t

0.44

Activations Density 1.254%

attends to negated or contrasting phrases from their preceding positive statements

No Comments

No Known Activations