INDEX

Explanations

verbs associated with pinning or tagging actions

oai_token-act-pair · gpt-3.5-turbo

phrases related to being constrained or categorized in some way

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

Exp

-0.80

 dece

-0.74

 Ether

-0.72

NEC

-0.68

 desert

-0.67

acter

-0.66

 hypocr

-0.66

ocr

-0.66

 atmosphere

-0.65

 profit

-0.65

POSITIVE LOGITS

 pinned

3.46

 pegged

2.06

 tagged

1.80

 tackled

1.59

 nailed

1.36

 framed

1.30

 chained

1.29

 flagged

1.27

 tied

1.25

 clipped

1.25

Activations Density 0.043%

verbs associated with pinning or tagging actions

phrases related to being constrained or categorized in some way

No Comments

No Known Activations