INDEX

Explanations

terms related to roles or positions in different contexts

oai_token-act-pair · gpt-3.5-turbo

references to astrological concepts or the term "roll."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 exhibit

-0.66

 Fourth

-0.64

 enterprise

-0.61

quickShipAvailable

-0.61

 Tsukuyomi

-0.61

 exhibits

-0.60

 impulse

-0.59

 Haiti

-0.58

 fashion

-0.57

 infinity

-0.56

POSITIVE LOGITS

rol

4.72

rology

2.19

rolog

2.06

rolled

1.61

rogen

1.46

roll

1.45

ros

1.38

rog

1.36

roller

1.34

rogens

1.29

Activations Density 0.008%

terms related to roles or positions in different contexts

references to astrological concepts or the term "roll."

No Comments

No Known Activations