INDEX

Explanations

specific combinations of letters or codes that seem to represent entities or resources

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

user turn indicators

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

KommentareTeilen

-0.56

 zależy

-0.47

 tričko

-0.45

 financieros

-0.43

ManagedBean

-0.42

 almohada

-0.42

HomeAsUp

-0.40

 względu

-0.39

 Mehrheit

-0.39

 télécommande

-0.39

POSITIVE LOGITS

ंदीखरीदारी

0.50

 SPIN

0.48

 ✭✭

0.47

hyrchwyd

0.47

 Exactos

0.44

 Shed

0.43

 Spin

0.43

 Rude

0.43

mut

0.42

 comput

0.41

Activations Density 0.007%

specific combinations of letters or codes that seem to represent entities or resources

user turn indicators

No Comments

No Known Activations