INDEX

Explanations

instances of reasoning, debate, and the presentation of opinions or arguments

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

decision, debate, opinion, truth, search

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

 which

-0.36

 those

-0.34

 перено

-0.33

 Vorg

-0.31

 interacting

-0.30

 entire

-0.30

 applying

-0.30

 Heiligen

-0.29

 suces

-0.29

 olev

-0.29

POSITIVE LOGITS

endphp

0.71

angliski

0.68

 estekak

0.64

majánló

0.64

localctx

0.63

ロウィン

0.62

 tartalomajánló

0.60

NameInMap

0.60

 パンチラ

0.60

 témoig

0.59

Activations Density 0.067%

instances of reasoning, debate, and the presentation of opinions or arguments

decision, debate, opinion, truth, search

No Comments

No Known Activations