INDEX

Explanations

attends to tokens that are marked as references or citations from earlier 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

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.16

1:0.23

2:0.09

3:0.06

4:0.10

5:0.12

6:0.06

7:0.14

Negative Logits

__(/*!

-0.39

 Krie

-0.36

esseract

-0.35

Loh

-0.35

+:+

-0.34

ieteur

-0.33

<h6>

-0.33

glLoadIdentity

-0.32

böz

-0.31

AndPassword

-0.31

POSITIVE LOGITS

Xna

0.42

tagHelperRunner

0.34

 Virginie

0.32

 planas

0.32

 oliveira

0.32

 kulak

0.32

 esternos

0.32

 Villarreal

0.31

 فريبيس

0.31

Carla

0.31

Activations Density 0.385%

attends to tokens that are marked as references or citations from earlier tokens

No Comments

No Known Activations