INDEX

Explanations

symbols and punctuation related to mathematical expressions and data structures

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

matrix notation

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

 contextLoads

-0.34

 inilah

-0.29

UserScript

-0.27

 TestBed

-0.27

Geplaatst

-0.26

 combineReducers

-0.26

Hentet

-0.25

 filósofo

-0.25

licherweise

-0.24

ĩa

-0.24

POSITIVE LOGITS

 StringTokenizer

0.58

thenReturn

0.58

FunctionFlags

0.57

 zwiſchen

0.57

enterOuterAlt

0.54

 autorytatywna

0.54

ViewFeatures

0.52

RenderAtEndOf

0.52

PreferredItem

0.52

arraycopy

0.52

Activations Density 0.063%

symbols and punctuation related to mathematical expressions and data structures

matrix notation

No Comments

No Known Activations