INDEX

Explanations

arbitrary symbols or characters that are not part of regular text, possibly indicating a coding error or corrupted data

oai_token-act-pair · gpt-3.5-turbo

instances of the character "Ŀ"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 unconscious

-0.84

 circulation

-0.79

 imperson

-0.76

 cradle

-0.75

 unborn

-0.71

 assemb

-0.69

 tackling

-0.69

 graveyard

-0.68

 likeness

-0.68

 casing

-0.68

POSITIVE LOGITS

VERTISEMENT

0.99

Â¯

0.91

ï¸

0.87

VII

0.84

ï¸ı

0.83

âĻ¦

0.83

Â°

0.82

âĸ

0.81

SourceFile

0.80

âĢł

0.79

Activations Density 0.232%

arbitrary symbols or characters that are not part of regular text, possibly indicating a coding error or corrupted data

instances of the character "Ŀ"

No Comments

No Known Activations