INDEX

Explanations

exclamatory and questioning phrases with strong emotions

oai_token-act-pair · gpt-3.5-turbo

punctuation marks, particularly exclamation marks and question marks

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Logged

-0.72

intent

-0.63

hunt

-0.63

Questions

-0.62

 WARN

-0.60

 ÃĹ

-0.60

PIN

-0.60

Submit

-0.59

Topic

-0.57

 FULL

-0.57

POSITIVE LOGITS

ooters

1.06

etc

0.85

iability

0.82

 respectively

0.74

itaire

0.73

 which

0.72

wark

0.71

igree

0.71

antha

0.70

iband

0.68

Activations Density 0.037%

exclamatory and questioning phrases with strong emotions

punctuation marks, particularly exclamation marks and question marks

No Comments

No Known Activations