INDEX

Explanations

references to historical and religious texts or 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_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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 interns

-0.87

 lockout

-0.81

 lasers

-0.79

 runway

-0.77

 unsafe

-0.76

 rookie

-0.76

 rollout

-0.76

 broadcast

-0.76

 backdoor

-0.75

 Intercept

-0.75

POSITIVE LOGITS

Ã¦

1.24

á¸

1.21

Ã»

1.20

ospels

1.18

anus

1.14

atha

1.12

Äģ

1.11

Ã¢

1.10

ibli

1.09

akh

1.09

Activations Density 0.344%

references to historical and religious texts or events

No Comments

No Known Activations