INDEX

Explanations

formal structures in the text

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

documentation headers

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

 lidé

-0.44

 Normdatei

-0.41

bcryptjs

-0.39

Autoritní

-0.36

 hanem

-0.36

今後も

-0.36

 doká

-0.35

 udrž

-0.35

 dagegen

-0.35

 kormány

-0.34

POSITIVE LOGITS

rawDesc

0.63

WriteTagHelper

0.57

principalTable

0.56

✭✭

0.55

клопе

0.52

########.

0.50

roides

0.50

 AssemblyCompany

0.50

 defaultstate

0.49

Activations Density 0.021%

formal structures in the text

documentation headers

No Comments

No Known Activations