INDEX

Explanations

phrases indicating choice or contrast between two options

oai_token-act-pair · gpt-3.5-turbo

references to the concept of choosing between alternatives or pairs

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

humans

-0.72

Psy

-0.68

ove

-0.67

Topics

-0.65

ships

-0.65

lux

-0.64

obin

-0.64

ories

-0.63

pir

-0.63

ARS

-0.62

POSITIVE LOGITS

worldly

0.86

 lobe

0.84

 equally

0.80

dayName

0.74

 glance

0.74

 baseman

0.74

 wart

0.72

ngth

0.69

 ingred

0.68

heastern

0.66

Activations Density 0.055%

phrases indicating choice or contrast between two options

references to the concept of choosing between alternatives or pairs

No Comments

No Known Activations