INDEX

Explanations

punctuation marks and dashes, particularly those indicating breaks or interruptions in text

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

sentence fragments after punctuation

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

 such

-0.35

 novios

-0.31

reg

-0.31

Grüsse

-0.31

max

-0.31

AssignableFrom

-0.30

such

-0.30

 completely

-0.30

image

-0.30

 Näch

-0.30

POSITIVE LOGITS

Aholisi

0.77

 autorytatywna

0.66

niſſe

0.63

ſelf

0.60

iſchen

0.60

Autoritní

0.59

iſche

0.59

الحياه

0.59

 surla

0.59

 ویکی‌پدی

0.58

Activations Density 0.007%

punctuation marks and dashes, particularly those indicating breaks or interruptions in text

sentence fragments after punctuation

No Comments

No Known Activations