INDEX

Explanations

phrases related to conveying positive information or updates

oai_token-act-pair · gpt-3.5-turbo

phrases indicating good news

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

senal

-0.83

aval

-0.73

urat

-0.71

 occup

-0.71

arij

-0.69

 trained

-0.68

 solvent

-0.67

ONSORED

-0.67

 sediment

-0.67

apons

-0.63

POSITIVE LOGITS

worthiness

0.88

worthy

0.87

night

0.75

Catholic

0.74

 headline

0.74

reader

0.74

 headlines

0.73

peak

0.73

mith

0.73

NEWS

0.71

Activations Density 0.033%

phrases related to conveying positive information or updates

phrases indicating good news

No Comments

No Known Activations