INDEX

Explanations

negative and offensive language and phrases

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

İĭ

-0.69

ãĥĥãĥĪ

-0.68

 Incarnation

-0.63

hof

-0.63

âĸ¬âĸ¬

-0.63

ĪĴ

-0.63

 STATES

-0.62

 separatist

-0.61

 presumptive

-0.60

naires

-0.59

POSITIVE LOGITS

atial

1.24

ooky

1.23

aghetti

1.19

rawl

1.13

rit

1.13

ruce

1.11

anish

1.10

acer

1.05

rocket

1.04

indle

1.03

Activations Density 0.403%

negative and offensive language and phrases

No Comments

No Known Activations