INDEX

Explanations

phrases and terms related to assistance and support in various contexts

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

consequence for better outcomes

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_24/width_131k/average_l0_97

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.24.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

躇

-0.60

 Meksiku

-0.58

gdx

-0.58

Efq

-0.57

ScopeManager

-0.52

 Anſ

-0.50

bootstrapcdn

-0.49

 disambiguazione

-0.49

 ―――――

-0.48

ENODEV

-0.47

POSITIVE LOGITS

 ensure

0.62

 ensures

0.60

 asegurarse

0.47

 avoid

0.47

 better

0.46

 ensuring

0.45

 prevent

0.45

 your

0.45

you

0.43

 avoids

0.43

Activations Density 0.018%

phrases and terms related to assistance and support in various contexts

consequence for better outcomes

No Comments

No Known Activations