INDEX

Explanations

locations or addresses

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

 Ninth

-0.82

 Tenth

-0.76

 Aires

-0.75

Boo

-0.74

 Goldberg

-0.71

 Parenthood

-0.71

 Piercing

-0.70

 Singer

-0.70

 Versus

-0.69

 Brill

-0.69

POSITIVE LOGITS

1.58

info

1.27

__

1.25

_-_

1.21

/)

1.15

reports

1.14

bg

1.09

stats

1.08

fc

1.08

/,

1.08

Activations Density 2.252%

locations or addresses

No Comments

No Known Activations