INDEX

Explanations

specific programming constructs and data types

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

graphics precision qualifiers

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

 Seuss

-0.49

 Thule

-0.47

 IBOutlet

-0.46

__*/

-0.46

poles

-0.46

bent

-0.45

 Versa

-0.43

 שוליים

-0.43

 invokingState

-0.43

 Senses

-0.42

POSITIVE LOGITS

ImageContext

0.55



0.51

 المعيارى

0.44

 Putri

0.43

氘

0.42

 contextLoads

0.41

rungsseite

0.41

 chi̍t

0.40

 congreso

0.40

gjenge

0.39

Activations Density 0.001%

specific programming constructs and data types

graphics precision qualifiers

No Comments

No Known Activations