INDEX

Explanations

sequences of text that indicate presence or specifications of data-related elements

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

numbers following section titles

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

dcmtk

-0.34

 enamorado

-0.32

 maravilla

-0.32

 dictadura

-0.32

 antigüedad

-0.30

 onOptions

-0.30

 complejidad

-0.28

 main

-0.28

élimin

-0.28

 funções

-0.27

POSITIVE LOGITS

الحياه

0.69

tvguidetime

0.68

 nakalista

0.67

 pinulongan

0.67

Tikang

0.62

 témoig

0.61

setVerticalGroup

0.60

RegressionTest

0.60

 invokingState

0.60

<unused30>

0.57

Activations Density 0.009%

sequences of text that indicate presence or specifications of data-related elements

numbers following section titles

No Comments

No Known Activations