INDEX

Explanations

attends to the token "speaking" from tokens that refer to the subject of what is being mentioned or discussed later in the text

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

0:0.15

1:0.16

2:0.15

3:0.11

4:0.04

5:0.01

6:0.08

7:0.26

Negative Logits

GEBURTSDATUM

-0.44

ftagPool

-0.37

 JpaRepository

-0.32

تقاوى

-0.29

èdia

-0.29

 uintptr

-0.29

afficheront

-0.28

Phaser

-0.28

RTSC

-0.27

 protoimpl

-0.27

POSITIVE LOGITS

GHIJKLM

0.27

 referenties

0.25

 kuin

0.24

 futuros

0.24

 demain

0.24

 Geschäfts

0.24

 vastaan

0.23

\\

0.23

 course

0.23

 szár

0.23

Activations Density 0.061%

attends to the token "speaking" from tokens that refer to the subject of what is being mentioned or discussed later in the text

No Comments

No Known Activations