INDEX

Explanations

hyphenated numerical values

oai_token-act-pair · gpt-3.5-turbo

negation or negative contexts

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

 Mechdragon

-0.66

eele

-0.65

EW

-0.63

 Ober

-0.63

PsyNetMessage

-0.62

Ard

-0.61

IPM

-0.61

Ud

-0.60

 augmented

-0.60

 Moody

-0.60

POSITIVE LOGITS

dozen

1.18

few

1.07

pill

1.00

ccording

0.98

minute

0.96

century

0.95

compl

0.95

million

0.94

billion

0.94

week

0.92

Activations Density 0.029%

hyphenated numerical values

negation or negative contexts

No Comments

No Known Activations