INDEX

Explanations

terms related to quality and professional-grade specifications in various contexts

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

level or type descriptor

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

 speranza

-0.38

 getRule

-0.34

 szóci

-0.34

 blessé

-0.33

 olvides

-0.29

 féminin

-0.29

 burbu

-0.29

 diciendo

-0.28

 pere

-0.28

 magnétique

-0.28

POSITIVE LOGITS

grade

0.68

 grade

0.63

professionals

0.60

bootstrapcdn

0.59

 Normdatei

0.59

 للاسماء

0.59

 GRADE

0.58

quality

0.58

professional

0.57

 BoxFit

0.56

Activations Density 0.011%

terms related to quality and professional-grade specifications in various contexts

level or type descriptor

No Comments

No Known Activations