INDEX

Explanations

references to legal cases and courtroom-related vocabulary

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

thood

-0.86

Ò

-0.81

bg

-0.74

ceive

-0.74

perse

-0.74

ï¿½ï¿½ï¿½ï¿½

-0.73

icia

-0.72

leeve

-0.71

imi

-0.70

aba

-0.70

POSITIVE LOGITS

oret

1.50

 latter

1.35

 biggest

1.27

resa

1.25

 remainder

1.21

 slightest

1.19

odore

1.18

 majority

1.17

 simplest

1.15

 vast

1.15

Activations Density 3.765%

references to legal cases and courtroom-related vocabulary

No Comments

No Known Activations