INDEX

Explanations

words related to compliance or adherence

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

ifter

-0.61

dar

-0.60

stall

-0.60

bin

-0.59

oufl

-0.58

 Monaco

-0.57

 Leone

-0.56

andel

-0.55

 Oblivion

-0.54

 Helsinki

-0.53

POSITIVE LOGITS

 thereto

1.45

to

1.05

itiz

0.85

ences

0.83

To

0.81

entious

0.79

 unto

0.78

itionally

0.78

ities

0.76

itive

0.74

Activations Density 5.093%

words related to compliance or adherence

No Comments

No Known Activations