INDEX

Explanations

attends to the last token in the sequence from tokens in the middle of 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

0:0.09

1:0.12

2:0.12

3:0.15

4:0.11

5:0.07

6:0.16

7:0.14

Negative Logits

Rohy

-0.39

 estimés

-0.39

UnsafeEnabled

-0.38

 صوتيه

-0.36

ChildScrollView

-0.36

viewDidLoad

-0.36

脚注の使い方

-0.35

 Infórmanos

-0.34

QtWidgets

-0.34

鰭

-0.33

POSITIVE LOGITS

IonicModule

0.23

RequestMethod

0.22

vábbi

0.20

IIA

0.19

 långt

0.19

Források

0.19

 whom

0.19

kA

0.19

 unrelated

0.19

يفة

0.19

Activations Density 0.371%

attends to the last token in the sequence from tokens in the middle of the sequence

No Comments

No Known Activations