INDEX

Explanations

the word "so"

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

 Mehran

-0.67

 wiser

-0.62

 Crowley

-0.60

 toast

-0.56

kh

-0.56

 ALEC

-0.55

 plotting

-0.54

scribe

-0.54

 Purg

-0.54

 pursu

-0.54

POSITIVE LOGITS

 many

1.03

undreds

0.91

many

0.88

aps

0.85

oru

0.84

bs

0.84

 much

0.78

azard

0.78

oooo

0.76

ickets

0.76

Activations Density 0.070%

the word "so"

No Comments

No Known Activations