INDEX

Explanations

proper nouns, likely related to individuals or locations

oai_token-act-pair · gpt-3.5-turbo

words and suffixes related to personal names and identities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 reckoning

-0.78

jriwal

-0.76

behind

-0.70

 pointers

-0.69

 professors

-0.67

 textbooks

-0.66

 poles

-0.65

 wiser

-0.65

 fractions

-0.64

 inclination

-0.64

POSITIVE LOGITS

tein

0.94

uve

0.92

Ã³n

0.91

cia

0.91

chio

0.89

oux

0.88

aux

0.85

ius

0.84

zza

0.84

Ã£o

0.83

Activations Density 0.263%

proper nouns, likely related to individuals or locations

words and suffixes related to personal names and identities

No Comments

No Known Activations