INDEX

Explanations

ways to solve problems or achieve goals

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

interstitial

-0.85

agos

-0.76

mite

-0.73

awn

-0.70

atched

-0.70

 earthqu

-0.70

Crystal

-0.70

attery

-0.69

cious

-0.69

isSpecialOrderable

-0.69

POSITIVE LOGITS

how

1.25

 ways

1.10

why

1.08

WHY

0.98

how

0.93

HOW

0.92

 what

0.91

 whats

0.87

 exactly

0.85

 whether

0.83

Activations Density 0.030%

ways to solve problems or achieve goals

No Comments

No Known Activations