INDEX

Explanations

patterns related to asking questions or seeking information

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

rising

-0.71

soDeliveryDate

-0.70

 Kinnikuman

-0.69

 Stru

-0.67

soever

-0.66

Lag

-0.64

 Rising

-0.63

 Compass

-0.63

abiding

-0.62

packing

-0.61

POSITIVE LOGITS

 emulate

1.33

 hear

1.23

 maximize

1.18

 keep

1.12

 recreate

1.10

 avoid

1.09

 know

1.08

see

1.08

 capitalize

1.08

 pursue

1.08

Activations Density 0.636%

patterns related to asking questions or seeking information

No Comments

No Known Activations