INDEX

Explanations

attends to specific variable names or types from empty or comment tokens that may follow them

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

0:0.07

1:0.19

2:0.12

3:0.13

4:0.13

5:0.07

6:0.17

7:0.08

Negative Logits

expandindo

-0.37

ViewFeatures

-0.26

 περ

-0.24

reps

-0.24

("");

-0.24

Portály

-0.23

 Wikimedijinoj

-0.23

 '\\;'

-0.22

WillAppear

-0.22

POSITIVE LOGITS

OGND

0.35

adpleegd

0.35

erdings

0.33

tahui

0.32

 whole

0.32

mologie

0.31

дожник

0.30

ngdoc

0.30

ONOM

0.30

bibinfo

0.30

Activations Density 0.000%

attends to specific variable names or types from empty or comment tokens that may follow them

No Comments

No Known Activations