INDEX

Explanations

weapon-related terms like 'gun', 'rifle', and 'knife'

oai_token-act-pair · gpt-3.5-turbo

nouns related to weapons and objects associated with violence or crime

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

 Helpful

-0.85

 ______

-0.65

 Flavoring

-0.65

erenn

-0.63

ilaterally

-0.62

LLP

-0.60

 sqor

-0.60

EStreamFrame

-0.60

___

-0.59

POSITIVE LOGITS

 disappears

1.06

 wasn

1.05

 belonged

1.00

 belongs

1.00

 hasn

0.95

isn

0.95

was

0.94

 doesn

0.90

 appears

0.89

 explodes

0.88

Activations Density 0.527%

weapon-related terms like 'gun', 'rifle', and 'knife'

nouns related to weapons and objects associated with violence or crime

No Comments

No Known Activations