INDEX

Explanations

mathematical expressions and technical details

oai_token-act-pair · gpt-3.5-turbo

negative or low numerical values, particularly in mathematical or statistical contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ÂŃ

-0.72

bee

-0.63

 painstaking

-0.59

 Elder

-0.59

authent

-0.59

advertisement

-0.57

 authent

-0.57

DNA

-0.56

Bee

-0.56

blogs

-0.55

POSITIVE LOGITS

(-

3.73

(+

2.40

 (âĪĴ

2.27

(-

1.84

[-

1.70

(~

1.54

(/

1.52

(_

1.44

(<

1.43

(=

1.41

Activations Density 0.010%

mathematical expressions and technical details

negative or low numerical values, particularly in mathematical or statistical contexts

No Comments

No Known Activations