INDEX

Explanations

phrases that indicate varying degrees of moderation or qualification in opinions or descriptions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_16k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 HasFactory

-0.62

 enfans

-0.47

 amarillas

-0.46

 Inscrivez

-0.44

 Meksiku

-0.44

 posesión

-0.44

tiap

-0.43

 aveug

-0.43

 vectorielles

-0.40

 industriels

-0.40

POSITIVE LOGITS

bit

0.91

 slightly

0.81

 Somewhat

0.78

 Slightly

0.77

 somewhat

0.77

Bit

0.73

やや

0.72

Somewhat

0.71

BIT

0.71

 vähän

0.69

Activations Density 0.139%

phrases that indicate varying degrees of moderation or qualification in opinions or descriptions

No Comments

No Known Activations