INDEX

Explanations

numerical sequences in the format "1_9" or "1_10"

oai_token-act-pair · gpt-3.5-turbo

repeated placeholder tokens, indicating the presence of segments marked by specific formatting or structure rather than content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 6-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.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

 fences

-0.72

 boundaries

-0.69

 modelling

-0.69

scenes

-0.68

bent

-0.64

 aesthetics

-0.64

thinkable

-0.64

awaru

-0.64

leading

-0.64

 hygiene

-0.63

POSITIVE LOGITS

Password

1.30

1.14

 Corinthians

1.12

st

1.10

1.09

Â½

0.95

0.94

0.91

RM

0.86

Activations Density 0.052%

numerical sequences in the format "1_9" or "1_10"

repeated placeholder tokens, indicating the presence of segments marked by specific formatting or structure rather than content

No Comments

No Known Activations