INDEX

Explanations

attends to style-related tokens from input-related 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

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.10

1:0.14

2:0.09

3:0.08

4:0.11

5:0.02

6:0.15

7:0.27

Negative Logits

PutMapping

-0.32

opla

-0.30

NotBlank

-0.30

antle

-0.28

 createState

-0.28

renza

-0.27

Quellen

-0.27

enterOuterAlt

-0.27

кота

-0.27

Eck

-0.26

POSITIVE LOGITS

 Houſe

0.47

 Majefty

0.46

 Jefus

0.45

ſelves

0.44

 myſelf

0.44

 Diſ

0.43

 iſt

0.43

 pleaſure

0.42

 unſ

0.42

 houſe

0.42

Activations Density 0.001%

attends to style-related tokens from input-related tokens

No Comments

No Known Activations