INDEX

Explanations

references to spatial direction and orientation

oai_token-act-pair · gpt-4o-mini Triggered by @bot

directions and turns

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ualitas

-0.36

 programmation

-0.34

Année

-0.34

restre

-0.33

Program

-0.31

 stos

-0.31

子里

-0.31

 verbund

-0.31

 inhoud

-0.31

 erop

-0.31

POSITIVE LOGITS

HideFlags

0.64

SourceChecksum

0.63

 مرئيه

0.63

JspWriter

0.61

 الحره

0.60

GOTREF

0.60

RegistryLite

0.58

richTextPanel

0.58

ArgumentParser

0.57

betical

0.56

Activations Density 0.024%

references to spatial direction and orientation

directions and turns

No Comments

No Known Activations