INDEX

Explanations

attends to empty tokens or placeholders from tokens later in the sequence

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

0:0.28

1:0.27

2:0.09

3:0.06

4:0.05

5:0.04

6:0.06

7:0.12

Negative Logits

tagHelperRunner

-0.60

 lenker

-0.48

SourceChecksum

-0.47

 propOrder

-0.46

цездатний

-0.44

 صوتيه

-0.42

BeginContext

-0.41

:✨

-0.41

 autorytatywna

-0.41

Einzelnachweise

-0.40

POSITIVE LOGITS

ins

0.26

</em>

0.24

 Season

0.24

Season

0.22

 season

0.22

land

0.21

</u>

0.21

 Carol

0.21

Ga

0.21

Activations Density 0.298%

attends to empty tokens or placeholders from tokens later in the sequence

No Comments

No Known Activations