INDEX

Explanations

phrases directed at the reader with a varying overall tone

oai_token-act-pair · gpt-3.5-turbo

references to the second person, specifically addressing the reader or listener directly

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 6-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.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

ice

-0.73

assembly

-0.70

icy

-0.69

¿½

-0.67

ftime

-0.64

î

-0.63

 Samoa

-0.61

 airs

-0.60

UTC

-0.60

 Commerce

-0.59

POSITIVE LOGITS

're

1.52

've

1.27

'll

1.24

'd

1.07

 guys

1.00

tub

1.00

owe

0.91

 know

0.90

 gotta

0.88

RS

0.87

Activations Density 0.260%

phrases directed at the reader with a varying overall tone

references to the second person, specifically addressing the reader or listener directly

No Comments

No Known Activations