INDEX

Explanations

patterns or sequences of symbols, particularly those that are visually distinct or formatted in a unique way

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

New Auto-Interp

Top Features by Cosine Similarity

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

Configuration

google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71

Prompts (Dashboard)

36,864 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

 itemprop

-0.53

 menj

-0.52

Eff

-0.49

原神

-0.49

 quæ

-0.48

 beijo

-0.47

ʔ

-0.46

 acuario

-0.45

 Segen

-0.45

 Herrick

-0.44

POSITIVE LOGITS

................

0.88

Personendaten

0.74

 propOrder

0.66

CloseOperation

0.65

ConstraintMaker

0.65

saraba

0.63

')")

0.62

BeginContext

0.60

...............

0.58

..............

0.57

Activations Density 0.125%

patterns or sequences of symbols, particularly those that are visually distinct or formatted in a unique way

No Comments

No Known Activations