INDEX

Explanations

quantifiers and phrases that indicate amounts or counts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 17-gemmascope-res-65k

Configuration

google/gemma-scope-2b-pt-res/layer_17/width_65k/average_l0_125

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.17.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

 poffe

-0.90

脚注の使い方

-0.87

 pleaſure

-0.85

 caufe

-0.81

 cauſe

-0.77

 juſt

-0.77

 fubject

-0.77

 &___

-0.77

 feveral

-0.75

 whofe

-0.75

POSITIVE LOGITS

of

0.71

 among

0.69

 amongst

0.67

 litro

0.64

among

0.56

 sorella

0.54

 Amongst

0.53

 للمعارف

0.52

poň

0.51

aarrggbb

0.50

Activations Density 0.021%

quantifiers and phrases that indicate amounts or counts

No Comments

No Known Activations