INDEX

Explanations

the abbreviation "OT" with increasing levels of intensity, reaching its highest activation with "OT" at 10

oai_token-act-pair · gpt-3.5-turbo

instances of the token "OT," indicating a focus on out-of-target content or specific categories in a dataset

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

bler

-0.72

uckland

-0.68

 fixme

-0.68

ãĥ¼ãĥĨ

-0.67

plain

-0.62

mM

-0.62

fold

-0.61

 nature

-0.61

antha

-0.61

 versa

-0.60

POSITIVE LOGITS

assium

1.17

TL

1.02

OGR

1.00

ECH

0.97

atoes

0.89

ION

0.88

TE

0.88

TO

0.87

OT

0.86

YP

0.85

Activations Density 0.015%

the abbreviation "OT" with increasing levels of intensity, reaching its highest activation with "OT" at 10

instances of the token "OT," indicating a focus on out-of-target content or specific categories in a dataset

No Comments

No Known Activations