INDEX

Explanations

attends to specific token patterns from arbitrary tokens earlier 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 @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.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

0:0.03

1:0.05

2:0.08

3:0.23

4:0.10

5:0.08

6:0.27

7:0.12

Negative Logits

-0.42

-0.39

-0.36

-0.35

-0.34

↵

-0.34

in

-0.33

POSITIVE LOGITS

Amicalement

0.68

 Normdatei

0.68

BibitemShut

0.64

 betweenstory

0.64

FTFY

0.62

ContentAsync

0.62

 itſelf

0.60

tvguidetime

0.59

sizeCache

0.59

)*/

0.58

Activations Density 0.002%

attends to specific token patterns from arbitrary tokens earlier in the sequence

No Comments

No Known Activations