INDEX

Explanations

terms indicating negation or non-existence

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

 Tycoon

-0.98

 Franks

-0.76

ãĤ¼ãĤ¦ãĤ¹

-0.68

 Grind

-0.65

 hordes

-0.65

 Rooms

-0.64

Dug

-0.64

 Halls

-0.64

Kings

-0.64

 Spoon

-0.62

POSITIVE LOGITS

chal

1.20

stop

1.10

linear

1.07

etheless

1.07

verbal

1.05

epad

1.04

fiction

1.02

threatening

1.02

profit

1.01

zero

1.00

Activations Density 0.018%

terms indicating negation or non-existence

No Comments

No Known Activations