INDEX

Explanations

the word "OK" with a strong activation value

oai_token-act-pair · gpt-3.5-turbo

expressions indicating a sense of approval or acceptance

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

urther

-0.60

ience

-0.59

cum

-0.59

 guiActiveUn

-0.58

Shadow

-0.57

eries

-0.57

ensis

-0.55

cence

-0.55

rowth

-0.55

 latent

-0.54

POSITIVE LOGITS

OK

3.95

ok

2.70

 okay

2.54

OK

2.32

 alright

2.09

 Okay

1.81

Ok

1.78

Ok

1.53

Okay

1.47

 Alright

1.34

Activations Density 0.005%

the word "OK" with a strong activation value

expressions indicating a sense of approval or acceptance

No Comments

No Known Activations