INDEX

Explanations

quantitative measurements and peaks in data

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

peak and maximal levels/rates

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

fjspx

-0.40

ніципалі

-0.39

 sereia

-0.37

 Numerade

-0.37

 tradition

-0.35

FieldOffsetTable

-0.35

KommentareTeilen

-0.35

 noDo

-0.35

iskt

-0.34

anthes

-0.33

POSITIVE LOGITS

 peak

0.57

peak

0.57

Peak

0.56

 peaked

0.54

+#+

0.54

twimg

0.50

 Peak

0.49

 peaking

0.49

 maxime

0.48

MessageState

0.47

Activations Density 0.044%

quantitative measurements and peaks in data

peak and maximal levels/rates

No Comments

No Known Activations