INDEX

Explanations

internet links or references to online content

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ridges

-0.87

theless

-0.82

size

-0.82

fully

-0.81

lier

-0.78

tle

-0.78

TON

-0.75

itud

-0.74

lear

-0.73

egg

-0.73

POSITIVE LOGITS

Äĩ

1.18

ya

0.94

plom

0.84

emi

0.80

aga

0.79

yah

0.78

ye

0.78

qa

0.76

ennes

0.75

arbon

0.75

Activations Density 0.034%

internet links or references to online content

No Comments

No Known Activations