INDEX

Explanations

numbers such as quantities, percentages, and years

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

↵Âł

-0.64

-0.61

 wisely

-0.60

Helper

-0.56

dal

-0.56

laws

-0.56

Tid

-0.56

rets

-0.55

Johnny

-0.55

Joined

-0.55

POSITIVE LOGITS

 dozen

1.38

dozen

1.36

 hundred

1.16

 thousand

1.10

 million

1.01

 lakh

1.00

ousand

0.90

%"

0.88

 thousands

0.86

 billion

0.85

Activations Density 0.863%

numbers such as quantities, percentages, and years

No Comments

No Known Activations