INDEX

Explanations

elements related to legal or evidentiary language

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

code and legal phrases

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

ReusableCell

-0.49

IntoConstraints

-0.46

AndroidJUnit

-0.45

momix

-0.43

 NavController

-0.43

esModule

-0.43

EndGlobalSection

-0.42

 MaterialApp

-0.41

endphp

-0.40

xase

-0.40

POSITIVE LOGITS

uxxxx

0.46

})));

0.44

ništ

0.43

 vimos

0.42

"]').

0.40

_));

0.40

 recibieron

0.40

 publiek

0.40

 defaultstate

0.40

 Paglinawan

0.40

Activations Density 0.001%

elements related to legal or evidentiary language

code and legal phrases

No Comments

No Known Activations