INDEX

Explanations

phrases where something is evaluated, assessed, or considered critically

oai_token-act-pair · gpt-3.5-turbo

the word "regarded" and its associations, indicating evaluation or reputation

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

adra

-0.70

 purse

-0.69

ells

-0.65

erry

-0.64

 canoe

-0.62

hatt

-0.61

 rain

-0.61

 bedrooms

-0.60

cot

-0.60

 Banks

-0.60

POSITIVE LOGITS

 regarded

1.01

CLASSIFIED

0.96

lehem

0.87

ritical

0.85

enance

0.81

Īè

0.79

æĺ¯

0.78

ouble

0.77

entious

0.76

ESCO

0.73

Activations Density 0.004%

phrases where something is evaluated, assessed, or considered critically

the word "regarded" and its associations, indicating evaluation or reputation

No Comments

No Known Activations