INDEX

Explanations

attends to tokens marked with parentheses from paired tokens that have a specific pattern related to legal codes or sections

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

0:0.27

1:0.20

2:0.16

3:0.06

4:0.07

5:0.06

6:0.04

7:0.09

Negative Logits

Personendaten

-0.35

帖最后由

-0.34

FunctionFlags

-0.33

bewerken

-0.33

haikusbot

-0.31

 تانيه

-0.31

HandleFunc

-0.30

Спасылкі

-0.30

ODO

-0.30

subpackage

-0.29

POSITIVE LOGITS

談社

0.30

kijken

0.28

pañas

0.27

QtWidgets

0.24

 supernova

0.24

Mě

0.24

ikä

0.24

pable

0.24

 okazji

0.24

posix

0.23

Activations Density 0.522%

attends to tokens marked with parentheses from paired tokens that have a specific pattern related to legal codes or sections

No Comments

No Known Activations