INDEX

Explanations

phrases related to concepts or entities being unquestionably true or confirmed

oai_token-act-pair · gpt-3.5-turbo

terms related to disputes or claims of authenticity and validity

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 stocking

-0.78

ikarp

-0.74

 cyan

-0.73

onse

-0.73

isson

-0.70

 newsletters

-0.69

 planner

-0.68

 tailored

-0.67

ollen

-0.66

 butterflies

-0.65

POSITIVE LOGITS

puted

1.31

uable

0.95

putable

0.82

teen

0.79

 Legends

0.73

uably

0.72

Fu

0.71

payer

0.71

enged

0.69

ulous

0.69

Activations Density 0.022%

phrases related to concepts or entities being unquestionably true or confirmed

terms related to disputes or claims of authenticity and validity

No Comments

No Known Activations