INDEX

Explanations

phrases related to durations of time, especially when referring to the past

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

 redistributed

-0.66

 Ahead

-0.62

 Seat

-0.58

 Cutting

-0.57

iru

-0.56

MSN

-0.56

ollo

-0.53

Collect

-0.53

},{"

-0.52

 Collector

-0.52

POSITIVE LOGITS

 millennia

1.22

 decades

1.21

 weeks

1.21

 years

1.20

 months

1.20

decade

1.18

 centuries

1.11

months

1.06

 hours

1.05

 periods

1.03

Activations Density 3.983%

phrases related to durations of time, especially when referring to the past

No Comments

No Known Activations