INDEX

Explanations

attends to the token appearing at the start of a code block from its associated ending token later 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.10

1:0.17

2:0.11

3:0.12

4:0.14

5:0.06

6:0.13

7:0.13

Negative Logits

RenderAtEndOf

-0.37

astéroïdes

-0.35

StructEnd

-0.35

 myſelf

-0.35

AsUp

-0.34

AnchorTagHelper

-0.34

 createSprite

-0.34

ftagPool

-0.33

 itſelf

-0.32

 оригіналу

-0.32

POSITIVE LOGITS

("")]

0.31

0.30

 parallel

0.28

 Parallel

0.26

jdzie

0.26

PARSE

0.26

Sh

0.26

higer

0.25

lopp

0.25

rigation

0.24

Activations Density 0.021%

attends to the token appearing at the start of a code block from its associated ending token later in the sequence

No Comments

No Known Activations