INDEX

Explanations

sentences with positive sentiment, expressions of gratitude, and requests for feedback

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

 instinct

-0.80

 questioning

-0.80

 elevated

-0.78

 rallying

-0.77

 imperson

-0.75

 tricked

-0.74

 undermining

-0.74

 escaping

-0.73

 raiding

-0.73

ensibly

-0.73

POSITIVE LOGITS

 Lastly

1.63

<|endoftext|>

1.60

 Additionally

1.42

 Alternatively

1.39

 Also

1.38

 Anyway

1.30

 Finally

1.24

 Please

1.22

 Enjoy

1.21

 Otherwise

1.19

Activations Density 4.496%

sentences with positive sentiment, expressions of gratitude, and requests for feedback

No Comments

No Known Activations