INDEX

Explanations

words related to types or categories

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

edia

-0.69

Lak

-0.67

 VIDEOS

-0.66

 Bakr

-0.61

mins

-0.61

UNCH

-0.60

eor

-0.60

 Bots

-0.60

ORY

-0.59

iets

-0.58

POSITIVE LOGITS

face

1.08

etting

1.04

casting

1.01

faces

0.96

lier

0.89

ahead

0.88

aldehyde

0.83

ifier

0.81

cast

0.80

alias

0.79

Activations Density 4.125%

words related to types or categories

No Comments

No Known Activations