INDEX

Explanations

attends to specific keywords or identifiers from earlier tokens in the sequence

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

0:0.08

1:0.19

2:0.06

3:0.04

4:0.14

5:0.37

6:0.05

7:0.04

Negative Logits

 photolibrary

-0.54

 myſelf

-0.52

tanleria

-0.51

IntoConstraints

-0.48

 ―――――

-0.48

 doubtnut

-0.47

SourceChecksum

-0.47

saraba

-0.47

InjectAttribute

-0.47

 therefrom

-0.47

POSITIVE LOGITS

<eos>

0.32

0.31

bric

0.29

is

0.29

ple

0.28

 host

0.27

urs

0.27

tu

0.27

0.26

 εγ

0.26

Activations Density 0.002%

attends to specific keywords or identifiers from earlier tokens in the sequence

No Comments

No Known Activations