INDEX

Explanations

phrases that indicate user engagement and interaction with content or process

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

reading and watching experiences

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



-0.61

 sanitarias

-0.55

 étrangère

-0.50

 propOrder

-0.48

 EconPapers

-0.47

ftagPool

-0.45

 policiales

-0.44

 étrangères

-0.44

Dichloropropane

-0.43

medriver

-0.43

POSITIVE LOGITS

 disambiguazione

0.38

 اطلع

0.37

IUrlHelper

0.37

 الاطلاع

0.35

 impression

0.35

 readers

0.34

 Lihat

0.33

 senses

0.33

parseFrom

0.33

 heard

0.32

Activations Density 0.102%

phrases that indicate user engagement and interaction with content or process

reading and watching experiences

No Comments

No Known Activations