INDEX

Explanations

years mentioned, especially when they are specified in a numerical form

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

 bluff

-0.62

 footh

-0.62

ktop

-0.62

 zipper

-0.60

 textual

-0.60

 cumbers

-0.59

 intermedi

-0.59

 sympath

-0.59

omething

-0.59

ogle

-0.58

POSITIVE LOGITS

long

1.33

nings

1.11

ning

1.03

book

1.01

ago

0.96

Ago

0.92

lies

0.91

 olds

0.90

iversary

0.88

glass

0.88

Activations Density 0.432%

years mentioned, especially when they are specified in a numerical form

No Comments

No Known Activations