INDEX

Explanations

specific phrases and words associated with communication and interactions

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

of followed by words like called, HMS, saying, record, prayer

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

 Romains

-0.46

 gräns

-0.41

 aplicable

-0.41

 Grecs

-0.41

 naturen

-0.39

 productivo

-0.38

 gardes

-0.36

 activer

-0.35

 områ

-0.35

 emperador

-0.35

POSITIVE LOGITS

sizeCache

0.75

 deſſen

0.69

iſchen

0.68

ſicht

0.68

 للمعارف

0.68

uxxxx

0.67

󠀠

0.66

 zuſammen

0.66

 queſto

0.66

 ſeiner

0.64

Activations Density 0.123%

specific phrases and words associated with communication and interactions

of followed by words like called, HMS, saying, record, prayer

No Comments

No Known Activations