INDEX

Explanations

phrases with the structure "Being a [noun]"

oai_token-act-pair · gpt-3.5-turbo

phrases that describe attributes or experiences related to identity or roles

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

 rounding

-0.66

 easing

-0.65

 awarding

-0.64

eren

-0.60

 Erie

-0.60

 intrigue

-0.60

 veins

-0.60

Wen

-0.59

urrence

-0.58

 Munich

-0.58

POSITIVE LOGITS

angelo

0.82

 Yourself

0.81

 oneself

0.78

ophobic

0.74

cribed

0.73

ovable

0.72

omorphic

0.72

thood

0.71

ammed

0.71

uned

0.70

Activations Density 0.323%

phrases with the structure "Being a [noun]"

phrases that describe attributes or experiences related to identity or roles

No Comments

No Known Activations