INDEX

Explanations

phrases related to addressing or acknowledging an audience

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.7.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 deterior

-0.76

 ordinance

-0.74

staking

-0.71

 euth

-0.67

 clauses

-0.66

 outweigh

-0.66

 destroys

-0.65

 lapt

-0.65

 withstand

-0.65

 excess

-0.65

POSITIVE LOGITS

 Fellow

0.82

Introdu

0.81

 welcome

0.77

yip

0.73

Exc

0.72

 Welcome

0.70

SEE

0.68

 Reader

0.67

Hello

0.67

Welcome

0.67

Activations Density 0.067%

phrases related to addressing or acknowledging an audience

No Comments

No Known Activations