INDEX

Explanations

apologies and expressions of hope or expectation

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

esp

-0.67

UFF

-0.61

 allegedly

-0.61

typically

-0.60

Downloadha

-0.59

 supposedly

-0.58

ãģ®éŃĶ

-0.57

usually

-0.57

inia

-0.56

 purportedly

-0.55

POSITIVE LOGITS

 someday

1.45

soon

0.89

 somew

0.82

 sooner

0.81

 eventually

0.80

 recons

0.77

 soon

0.75

 tomorrow

0.75

 hereafter

0.75

 sometime

0.74

Activations Density 1.962%

apologies and expressions of hope or expectation

No Comments

No Known Activations