INDEX

Explanations

long words that are not commonly used

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

hyde

-0.82

PORT

-0.73

Ô

-0.71

 Abyss

-0.70

 Hendricks

-0.69

 Mine

-0.68

 Millennium

-0.68

 Master

-0.67

 Fargo

-0.65

 Luck

-0.64

POSITIVE LOGITS

inging

1.08

akings

1.03

iddling

0.99

inged

0.99

idd

0.98

iper

0.96

ogging

0.93

Ð¾

0.92

ittance

0.90

tro

0.90

Activations Density 1.517%

long words that are not commonly used

No Comments

No Known Activations