INDEX

Explanations

numeric values within parentheses

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

 blot

-0.80

 impe

-0.73

 depress

-0.73

 drift

-0.72

 incons

-0.70

 baffled

-0.69

 jelly

-0.69

 swoop

-0.69

 impeachment

-0.68

 wrinkles

-0.68

POSITIVE LOGITS

formerly

1.25

aka

1.23

UGC

1.20

sic

1.11

which

1.09

NAS

1.08

also

1.06

NYSE

1.06

SG

1.06

including

1.05

Activations Density 0.269%

numeric values within parentheses

No Comments

No Known Activations