INDEX

Explanations

attends to specific token patterns, particularly those related to technical terms or symbols, from surrounding tokens

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

0:0.07

1:0.08

2:0.53

3:0.11

4:0.05

5:0.03

6:0.04

7:0.06

Negative Logits

AndEndTag

-0.45

 utafitiHapana

-0.37

RenderAtEndOf

-0.33

PathVariable

-0.31

 lenker

-0.28

 NgModule

-0.28

UnusedPrivate

-0.28

ิลปะ

-0.28

 poule

-0.28

 groupBox

-0.27

POSITIVE LOGITS

 nahilalakip

0.34

новниш

0.28

kuuta

0.27

()]

0.27

Tikang

0.26

crombie

0.25

Javadoc

0.25

($(

0.25

 isNameExpr

0.24

Скачать

0.24

Activations Density 0.292%

attends to specific token patterns, particularly those related to technical terms or symbols, from surrounding tokens

No Comments

No Known Activations