INDEX

Explanations

concepts related to differentiation and classification

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

distinguish true from false

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

WriteTagHelper

-0.55

 propOrder

-0.54

bootstrapcdn

-0.49

ChildScrollView

-0.46

الحياه

-0.43

AnchorStyles

-0.43

༘

-0.42

iotensin

-0.42

WebServlet

-0.41

CompleteListener

-0.41

POSITIVE LOGITS

 exceptions

0.43

standers

0.42

ThroughAttribute

0.41

ajemen

0.41

ніципалі

0.41

 virkelig

0.40

 boundaries

0.38

 hiér

0.38

 failures

0.38

 echten

0.37

Activations Density 0.056%

concepts related to differentiation and classification

distinguish true from false

No Comments

No Known Activations