INDEX

Explanations

news headlines indicating urgency or importance, often accompanied by a command to watch or read

oai_token-act-pair · gpt-3.5-turbo

urgent or emphatic directives

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

-0.80

 Hamp

-0.71

 Vish

-0.71

 Bhar

-0.69

gee

-0.68

bats

-0.66

itar

-0.66

gdala

-0.65

 Catalyst

-0.65

 Chak

-0.64

POSITIVE LOGITS

 WATCH

0.88

ered

0.82

ICLE

0.82

LECT

0.79

ARD

0.78

ering

0.77

nyder

0.77

 comply

0.77

 obey

0.74

ORTS

0.73

Activations Density 0.012%

news headlines indicating urgency or importance, often accompanied by a command to watch or read

urgent or emphatic directives

No Comments

No Known Activations