INDEX

Explanations

phrases that reformulate or simplify complex ideas

oai_token-act-pair · gpt-3.5-turbo

phrases indicating equivalence or clarification, particularly using the words "basically" and "essentially."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

alez

-0.80

Ey

-0.76

-0.70

TERN

-0.67

Others

-0.67

Tickets

-0.65

Recently

-0.63

icip

-0.63

eful

-0.60

Ced

-0.60

POSITIVE LOGITS

 worthless

0.82

 boils

0.78

 meaningless

0.74

 useless

0.71

 starve

0.68

nil

0.68

 copying

0.66

 indistinguishable

0.66

strate

0.65

 unchanged

0.65

Activations Density 0.038%

phrases that reformulate or simplify complex ideas

phrases indicating equivalence or clarification, particularly using the words "basically" and "essentially."

No Comments

No Known Activations