INDEX

Explanations

attends to conjunction tokens from tokens that indicate positional roles or responsibilities

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

0:0.10

1:0.11

2:0.11

3:0.08

4:0.12

5:0.04

6:0.15

7:0.26

Negative Logits

apos

-0.29

lemmer

-0.27

henswürdigkeiten

-0.27

彿

-0.26

="#"><

-0.25

he

-0.25

 onCreate

-0.25

 />);

-0.25

Carriera

-0.24

valget

-0.24

POSITIVE LOGITS

 themſelves

0.29

 eorum

0.29

UnsafeEnabled

0.29

 Monfieur

0.29

<bos>

0.29

 ainfi

0.29

 Mounts

0.28

 Metropolitana

0.28

Efq

0.27

 Böyle

0.27

Activations Density 0.243%

attends to conjunction tokens from tokens that indicate positional roles or responsibilities

No Comments

No Known Activations