INDEX

Explanations

terms related to detection and monitoring

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

anan

-0.77

 Holiday

-0.70

 presided

-0.67

 enjoyed

-0.66

orrow

-0.65

rue

-0.64

onte

-0.63

 socialist

-0.63

acre

-0.62

 Hammond

-0.62

POSITIVE LOGITS

 detection

3.82

 Detection

2.97

 detecting

2.37

 detector

2.28

 detectors

2.16

 detect

2.11

 detected

1.93

Detect

1.61

 detects

1.61

 Detect

1.59

Activations Density 0.015%

terms related to detection and monitoring

No Comments

No Known Activations