INDEX

Explanations

TV show titles

oai_token-act-pair · gpt-3.5-turbo

mentions of television shows and franchises

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 favor

-0.87

 exper

-0.86

 accrued

-0.85

 conflicted

-0.84

 mandate

-0.84

 descriptor

-0.82

 affected

-0.81

 impacts

-0.81

 resulting

-0.81

 mistakenly

-0.80

POSITIVE LOGITS

Pont

1.59

Sold

1.55

Bron

1.50

Street

1.49

Soul

1.48

Mar

1.47

License

1.46

Virgin

1.46

Sil

1.46

Silver

1.45

Activations Density 0.186%

TV show titles

mentions of television shows and franchises

No Comments

No Known Activations