INDEX

Explanations

attends to tokens marked "ac" from various numerical pairs

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.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

0:0.73

1:0.01

2:0.02

3:0.01

4:0.09

5:0.02

6:0.01

7:0.06

Negative Logits

BnF

-0.34



-0.34

ΗΣ

-0.30

Portale

-0.29

lgari

-0.28

piecze

-0.28

 promoc

-0.28

 Revi

-0.28

spaceBetween

-0.27

 defensively

-0.27

POSITIVE LOGITS

saraba

0.32

 wikipagina

0.32

 TestBed

0.31

ッケージ

0.31

 دریافت‌شده

0.31

 שוליים

0.31

Autoritní

0.30

Tembelea

0.29

 didik

0.29

 bezeichneter

0.29

Activations Density 0.015%

attends to tokens marked "ac" from various numerical pairs

No Comments

No Known Activations