INDEX

Explanations

phrases denoting a significant event for the first time

oai_token-act-pair · gpt-3.5-turbo

references to "first time" occurrences in history

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

$$$$

-0.62

ona

-0.60

dan

-0.58

 Gian

-0.57

_{

-0.56

mot

-0.56

hao

-0.55

ahn

-0.55

mo

-0.54

iltr

-0.54

POSITIVE LOGITS

 history

1.10

history

0.88

clusively

0.88

 existence

0.87

 awhile

0.85

 entirety

0.83

 succession

0.79

 History

0.76

ever

0.75

 decades

0.74

Activations Density 0.157%

phrases denoting a significant event for the first time

references to "first time" occurrences in history

No Comments

No Known Activations