INDEX

Explanations

proper nouns

oai_token-act-pair · gpt-3.5-turbo

specific letters and letter combinations that frequently appear in the text

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

²¾

-0.66

ãĥĳ

-0.63

·

-0.62

Ø¯

-0.62

ĩ

-0.60

«

-0.59

Ship

-0.58

Ĳ

-0.57

ĸ

-0.57

Bridge

-0.57

POSITIVE LOGITS

.—

0.62

 ðŁĺ

0.60

 ðŁĻĤ

0.57

 supers

0.53

 jail

0.50

 chant

0.49

 dominated

0.49

 ASAP

0.47

cham

0.47

 indefinitely

0.47

Activations Density 1.489%

proper nouns

specific letters and letter combinations that frequently appear in the text

No Comments

No Known Activations