INDEX

Explanations

mentions of a person named Rebecca, potentially with associated positive or negative context

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

ocene

-0.92

actic

-0.88

culosis

-0.88

axies

-0.83

ciating

-0.81

poral

-0.81

onymous

-0.80

oday

-0.79

uchin

-0.78

llular

-0.77

POSITIVE LOGITS

 Lynn

0.96

Bee

0.93

 Anne

0.91

 Louise

0.88

 Lopez

0.84

issance

0.84

 Kate

0.83

Anne

0.81

 Merkel

0.81

 Marie

0.81

Activations Density 0.032%

mentions of a person named Rebecca, potentially with associated positive or negative context

No Comments

No Known Activations