INDEX

Explanations

words indicating logical conclusions or cause and effect relationships

oai_token-act-pair · gpt-3.5-turbo

the word "therefore" and its synonyms or related transitional phrases indicating causation or conclusion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

•Causality

therefore + hence, thus

No Comments

Negative Logits

 Aman

-0.72

 Niet

-0.69

 Townsend

-0.67

Sov

-0.66

 Omaha

-0.66

 Paddock

-0.65

Ack

-0.65

 Siber

-0.65

 Cassandra

-0.64

 Fukushima

-0.63

POSITIVE LOGITS

untarily

0.87

forth

0.77

pired

0.75

ãĥĩãĤ£

0.75

deen

0.75

 proceeded

0.73

 undergo

0.71

received

0.71

 expressly

0.71

 opted

0.70

Activations Density 0.063%

words indicating logical conclusions or cause and effect relationships

the word "therefore" and its synonyms or related transitional phrases indicating causation or conclusion

No Comments

No Known Activations