INDEX

Explanations

scientific or technical terms related to laws, patents, organizations, and emotional concepts like poisons

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

vik

-0.73

DERR

-0.72

entimes

-0.70

erb

-0.67

lf

-0.66

biz

-0.64

irrel

-0.63

pmwiki

-0.61

ternity

-0.61

illy

-0.60

POSITIVE LOGITS

 respectively

1.15

 finalists

1.00

 listed

1.00

 vying

0.99

 composing

0.99

 simultaneously

0.94

 consecut

0.93

 mentioned

0.90

 combined

0.88

viz

0.87

Activations Density 0.297%

scientific or technical terms related to laws, patents, organizations, and emotional concepts like poisons

No Comments

No Known Activations