INDEX

Explanations

two-part words where the first part has a single letter and the second part is longer

oai_token-act-pair · gpt-3.5-turbo

sequences of letters that may correspond to various names, locations, or organizations

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

enegger

-0.52

jri

-0.49

 Leilan

-0.49

bom

-0.47

_.

-0.46

 psychiat

-0.46

 crib

-0.46

staking

-0.45

 compr

-0.45

 Vaugh

-0.44

POSITIVE LOGITS

 âĢº

0.46

LOG

0.45

GI

0.43

olitics

0.40

rad

0.39

 Brewing

0.39

Ph

0.39

LOD

0.39

olin

0.39

igon

0.39

Activations Density 0.652%

two-part words where the first part has a single letter and the second part is longer

sequences of letters that may correspond to various names, locations, or organizations

No Comments

No Known Activations