INDEX

Explanations

references to time or duration, particularly focusing on recent events

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

aunders

-0.76

ikan

-0.76

kaya

-0.71

cil

-0.70

udd

-0.69

gans

-0.68

 Franch

-0.67

keyes

-0.66

ordes

-0.65

 Swords

-0.65

POSITIVE LOGITS

 week

1.00

 month

0.97

 night

0.90

 year

0.88

 ditch

0.88

 rites

0.84

 evening

0.80

 gasp

0.79

 parting

0.78

 minute

0.77

Activations Density 0.076%

references to time or duration, particularly focusing on recent events

No Comments

No Known Activations