INDEX

Explanations

controversial or politically charged topics and events

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

 partName

-0.85

izons

-0.69

ciplinary

-0.68

Timer

-0.67

autions

-0.66

 anecd

-0.66

iple

-0.64

ishops

-0.63

 cautiously

-0.63

 concise

-0.62

POSITIVE LOGITS

 illegally

1.06

 allegedly

1.01

 pedoph

0.92

 raping

0.91

 sexually

0.89

 falsely

0.88

 purportedly

0.87

 homosexuality

0.87

 prostitutes

0.87

 illegal

0.86

Activations Density 13.590%

controversial or politically charged topics and events

No Comments

No Known Activations