INDEX

Explanations

instances of numbers and specific phrases like "with a grain of salt" and "the express understanding"

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

here

-0.77

estate

-0.74

adj

-0.73

sburg

-0.72

irable

-0.65

'/

-0.65

ombat

-0.64

ategories

-0.64

owship

-0.63

reality

-0.63

POSITIVE LOGITS

 caveat

1.21

 intention

1.08

 exception

1.03

 caveats

1.01

 flick

0.98

 impunity

0.97

 twist

0.96

 bang

0.93

 intent

0.91

 flourish

0.88

Activations Density 1.613%

instances of numbers and specific phrases like "with a grain of salt" and "the express understanding"

No Comments

No Known Activations