INDEX

Explanations

references to "value" and its variations, particularly in business or ethical contexts

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

value added

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_32/width_131k/average_l0_88

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.32.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

raw

-0.30

耍

-0.29

 Crow

-0.29

 вол

-0.29

 previous

-0.27

支

-0.27

Bu

-0.27

An

-0.26

Comp

-0.26

POSITIVE LOGITS

 houſe

0.73

 ब्रेकडाउन

0.66

 ChromeDriver

0.66

arakhand

0.63

 maximization

0.63

MessageOf

0.63

 betweenstory

0.62

 propositions

0.61

expandindo

0.60

proposition

0.59

Activations Density 0.010%

references to "value" and its variations, particularly in business or ethical contexts

value added

No Comments

No Known Activations