INDEX

Explanations

attends to document-related tokens from subsequent LaTeX code blocks

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

0:0.13

1:0.23

2:0.17

3:0.13

4:0.11

5:0.03

6:0.07

7:0.10

Negative Logits

runApp

-0.34

 EconPapers

-0.33

atzes

-0.32

æus

-0.30

SharedCtor

-0.30

 actionMode

-0.30

lapsingToolbar

-0.28

ichthys

-0.28

 Confederation

-0.27

➋

-0.27

POSITIVE LOGITS

GraphicsUnit

0.33

原始内容存档于

0.28

 berger

0.27

MLLoader

0.26

NewGuid

0.26

NDL

0.26

HttpStatusCode

0.25

StoryboardSegue

0.25

GUILayout

0.25

躇

0.25

Activations Density 0.012%

attends to document-related tokens from subsequent LaTeX code blocks

No Comments

No Known Activations