INDEX

Explanations

references to news outlets, specifically "The New York Times."

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

wcsstore

-0.76

backer

-0.72

vol

-0.72

olate

-0.72

milo

-0.70

ched

-0.70

ches

-0.70

cius

-0.69

alogue

-0.68

selection

-0.68

POSITIVE LOGITS

 City

1.39

 Mellon

1.05

 Harbor

1.02

 Times

0.98

 Cosmos

0.95

 Knicks

0.92

 Mets

0.92

 Jets

0.92

 State

0.90

 Yankees

0.89

Activations Density 0.935%

references to news outlets, specifically "The New York Times."

No Comments

No Known Activations