INDEX

Explanations

dates or times mentioned in the text

oai_token-act-pair · gpt-3.5-turbo

date references, specifically "Sep" followed by numerical values indicating days

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 blows

-0.71

 millions

-0.64

 careless

-0.61

 unprepared

-0.61

 bearer

-0.60

 unauthorized

-0.59

unf

-0.59

 teeth

-0.58

 liking

-0.57

 Mother

-0.55

POSITIVE LOGITS

arate

1.68

aration

1.58

arat

1.47

sis

1.22

arations

1.20

ar

1.04

hard

1.01

ul

0.99

hi

0.98

mented

0.97

Activations Density 0.030%

dates or times mentioned in the text

date references, specifically "Sep" followed by numerical values indicating days

No Comments

No Known Activations