INDEX

Explanations

the word "t"

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

 Reloaded

-0.64

 SetTextColor

-0.60

 Alleg

-0.60

Site

-0.59

 Dickinson

-0.55

 Houses

-0.55

 Penguin

-0.55

inia

-0.53

 Forums

-0.53

 Pengu

-0.53

POSITIVE LOGITS

 necessarily

0.98

 bother

0.86

urtles

0.86

actly

0.86

urtle

0.84

aper

0.83

ember

0.83

apest

0.82

ople

0.82

aylor

0.82

Activations Density 0.590%

the word "t"

No Comments

No Known Activations