INDEX

Explanations

phrases and words indicating causation or influence

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

 pigeon

-0.78

 Timber

-0.72

 Methodist

-0.66

conservancy

-0.64

 Alto

-0.63

 staples

-0.62

 quarters

-0.60

 Armenian

-0.60

Dud

-0.60

croft

-0.59

POSITIVE LOGITS

iveness

1.38

ual

1.24

uated

1.23

uating

1.16

uation

1.13

ually

1.11

uate

1.08

uates

1.07

ively

1.06

uality

1.00

Activations Density 0.028%

phrases and words indicating causation or influence

No Comments

No Known Activations