INDEX

Explanations

terms related to "RO" with a varying degree of emphasis through different levels of activation

oai_token-act-pair · gpt-3.5-turbo

mentions of the acronym "RO", likely indicating a focus on specific organizations or entities referenced in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Crit

-0.74

riad

-0.72

 Bundes

-0.72

 Kaufman

-0.71

 regards

-0.70

 Icelandic

-0.70

manship

-0.70

ãĥ¼ãĥĨãĤ£

-0.69

ãĤº

-0.69

 Norn

-0.69

POSITIVE LOGITS

bably

1.20

FILE

1.02

VERS

1.01

ptions

0.95

scill

0.94

spective

0.94

RO

0.92

LL

0.91

VOL

0.90

aches

0.90

Activations Density 0.007%

terms related to "RO" with a varying degree of emphasis through different levels of activation

mentions of the acronym "RO", likely indicating a focus on specific organizations or entities referenced in the text

No Comments

No Known Activations