INDEX

Explanations

code references to the current object or class context

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

user code blocks

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

 also

-0.44

ArrowToggle

-0.43

-0.39

too

-0.37

 with

-0.37

the

-0.36

 along

-0.36

now

-0.35

can

-0.35

 after

-0.35

POSITIVE LOGITS

Tembelea

0.66

 Administrativna

0.63

 propOrder

0.60

Personendaten

0.55

devamını

0.53

OGND

0.53

 Italijani

0.52

 للمعارف

0.52

BibitemShut

0.51

aarrggbb

0.50

Activations Density 0.018%

code references to the current object or class context

user code blocks

No Comments

No Known Activations