INDEX

Explanations

attends to tokens marked with specific keywords or data types from empty or generic 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

0:0.25

1:0.28

2:0.12

3:0.06

4:0.09

5:0.03

6:0.05

7:0.10

Negative Logits

aarrggbb

-0.35

 Normdatei

-0.32

żdy

-0.30

 Noodle

-0.30

 dreamstime

-0.30

 Comprometido

-0.30

addPreferredGap

-0.30

 lenker

-0.29

cautionary

-0.29

 Majefty

-0.29

POSITIVE LOGITS

 sendiri

0.27

 հղումներ

0.26

 EClass

0.25

uesia

0.24

 dAtA

0.23

lind

0.23

 Gang

0.23

CWE

0.22

featureID

0.22

es

0.22

Activations Density 6.484%

attends to tokens marked with specific keywords or data types from empty or generic tokens

No Comments

No Known Activations