INDEX

Explanations

prefixes like "Re" and "za" followed by a person's name and legal terms

oai_token-act-pair · gpt-3.5-turbo

complete sentences or statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 ambush

-0.81

 deity

-0.76

 tremend

-0.76

 barric

-0.75

 discreet

-0.74

 escort

-0.73

mur

-0.72

 bindings

-0.72

 instinct

-0.71

 tower

-0.70

POSITIVE LOGITS

 Additionally

1.33

 Overall

1.32

 Compared

1.28

 Similarly

1.24

 Moreover

1.23

 Conversely

1.21

 Furthermore

1.21

 Likewise

1.20

 However

1.18

 Therefore

1.15

Activations Density 0.596%

prefixes like "Re" and "za" followed by a person's name and legal terms

complete sentences or statements

No Comments

No Known Activations