INDEX

Explanations

numeric values associated with time and edits in a structured format

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

timestamp values

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

ThroughAttribute

-0.69

 صوتيه

-0.59

withIdentifier

-0.57

 NSCoder

-0.51

rungsseite

-0.50

 opérés

-0.48

 ostavi

-0.46

存于互联网档案馆

-0.45

 pinulongan

-0.42

หวัด

-0.42

POSITIVE LOGITS

 afternoon

0.46

afternoon

0.42

lifted

0.41

étation

0.40

 weighted

0.40

 CONSE

0.40

weighted

0.40

RTSC

0.39

 timestamp

0.39

gliche

0.39

Activations Density 0.044%

numeric values associated with time and edits in a structured format

timestamp values

No Comments

No Known Activations