INDEX

Explanations

words and phrases indicating significant values or headings in structured data formats

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

URL patterns from websites

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

findpost

-0.47

脚注の使い方

-0.44

GEBURTSDATUM

-0.42

 onOptions

-0.41

ArgsConstructor

-0.41

 NSCoder

-0.40

 '\\;'

-0.38

SPATH

-0.37

 "..\..\

-0.35

 positivos

-0.35

POSITIVE LOGITS

 betweenstory

0.53

Infant

0.47

 дописавши

0.47

 infant

0.45

 AssertionError

0.45

 Infant

0.43

SPECIFIC

0.43

 AssemblyCompany

0.41

infant

0.41

Mol

0.41

Activations Density 0.020%

words and phrases indicating significant values or headings in structured data formats

URL patterns from websites

No Comments

No Known Activations