INDEX

Explanations

phrases related to controversial topics and opinions

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

ngth

-0.78

luaj

-0.73

opez

-0.71

ividual

-0.68

aneers

-0.67

ifferent

-0.62

perty

-0.61

ebted

-0.60

isively

-0.60

vertisements

-0.60

POSITIVE LOGITS

 happening

0.91

 true

0.82

 understandable

0.81

quickShipAvailable

0.80

why

0.79

 compounded

0.79

 untrue

0.76

 blasphemy

0.74

 reassuring

0.73

SPONSORED

0.72

Activations Density 3.905%

phrases related to controversial topics and opinions

No Comments

No Known Activations