INDEX

Explanations

various forms of instructions and directives in the text

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

instructions and directions

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

slidesToShow

-0.43

oliber

-0.41

spesies

-0.40

󠁴

-0.38

UrlResolution

-0.37

TextBoxColumn

-0.37

actéris

-0.36

#+#

-0.36

 igenom

-0.36

BindView

-0.35

POSITIVE LOGITS

 instructions

1.30

instructions

1.17

 Instructions

1.16

Instructions

1.09

 INSTRUCTIONS

1.05

INSTRUCTIONS

0.95

 instruction

0.95

 directions

0.93

 instrucciones

0.93

 Directions

0.88

Activations Density 0.011%

various forms of instructions and directives in the text

instructions and directions

No Comments

No Known Activations