INDEX

Explanations

components related to names and identifiers in various contexts

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

code structure markers

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

 EconPapers

-0.73

olverine

-0.54



-0.51

⤹

-0.51

 uitz

-0.48

 sapat

-0.48

 tighten

-0.47

URIDAD

-0.47

 belangrij

-0.46

 století

-0.46

POSITIVE LOGITS

 BorderSide

0.54

 TextAlign

0.47

endregion

0.47

 NSCoder

0.40

 ModelExpression

0.39

ThroughAttribute

0.39

 дописавши

0.37

endmodule

0.37

fillType

0.36

archiviato

0.35

Activations Density 0.142%

components related to names and identifiers in various contexts

code structure markers

No Comments

No Known Activations