INDEX

Explanations

dates or time periods when something was conducted or researched

oai_token-act-pair · gpt-3.5-turbo

instances of the word "conducted" related to surveys or polls

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

inventoryQuantity

-0.72

å£«

-0.66

advertisement

-0.61

ĪĴ

-0.59

drop

-0.58

rising

-0.58

ingly

-0.58

mere

-0.58

="#

-0.58

 passage

-0.58

POSITIVE LOGITS

 jointly

0.97

 collabor

0.92

 concurrently

0.87

 outdoors

0.87

 anonymously

0.85

 indoors

0.77

 exclusively

0.75

 during

0.74

 separately

0.73

 undercover

0.73

Activations Density 0.132%

dates or time periods when something was conducted or researched

instances of the word "conducted" related to surveys or polls

No Comments

No Known Activations