INDEX

Explanations

phrases indicating the presence or lack of specific content or information (e.g., contained, no, contains)

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

bell

-0.84

doms

-0.77

aver

-0.69

camp

-0.67

pal

-0.66

hat

-0.65

aqu

-0.63

Sport

-0.63

 Sabha

-0.61

itus

-0.61

POSITIVE LOGITS

 prominently

0.89

 elements

0.85

 several

0.82

 lots

0.80

 plenty

0.80

 provisions

0.80

 dozens

0.76

 multiple

0.76

 considerable

0.75

 extensive

0.73

Activations Density 1.140%

phrases indicating the presence or lack of specific content or information (e.g., contained, no, contains)

No Comments

No Known Activations