INDEX

Explanations

descriptors of people, especially those that suggest age, condition, or reputation

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

adjective followed by noun

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

 definiert

-0.39

 primeiras

-0.39

 dále

-0.39

 continúa

-0.38

 asegurarse

-0.38

 definição

-0.38

 continues

-0.37

 demuestra

-0.36

 continuará

-0.36

 podrás

-0.35

POSITIVE LOGITS

$_"

0.91

contentLoaded

0.87

 Савезне

0.83

 Meksiku

0.78

 betweenstory

0.75

 للمعارف

0.75

 queſta

0.74

 firebaseConfig

0.72

MLLoader

0.72

 незавершена

0.71

Activations Density 0.068%

descriptors of people, especially those that suggest age, condition, or reputation

adjective followed by noun

No Comments

No Known Activations