INDEX

Explanations

references to web navigation elements and content structure

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

class name or start of turn

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

 Brü

-0.38

we

-0.38

 care

-0.36

 defined

-0.35

Sti

-0.33

Bü

-0.33

 political

-0.33

stalt

-0.31

 stuff

-0.31

 resource

-0.31

POSITIVE LOGITS

RegressionTest

0.79

TypedDataSet

0.79

LookAnd

0.78

 &___

0.68

ImageContext

0.68

AnchorStyles

0.65

 EconPapers

0.64

balleur

0.62

CppMethod

0.62

xygen

0.62

Activations Density 0.026%

references to web navigation elements and content structure

class name or start of turn

No Comments

No Known Activations