INDEX

Explanations

phrases related to advice or warnings

oai_token-act-pair · gpt-3.5-turbo

phrases indicating potential hazards or dangers

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 UNCLASSIFIED

-0.81

 thereafter

-0.76

 afterwards

-0.75

 continuity

-0.74

 stated

-0.74

emort

-0.72

 substantive

-0.72

 objectives

-0.68

 secondly

-0.67

 afterward

-0.66

POSITIVE LOGITS

Researchers

0.90

Scientists

0.83

toggle

0.82

utterstock

0.79

 Researchers

0.78

 Guinness

0.77

hello

0.76

 Nielsen

0.76

 Shutterstock

0.74

Redditor

0.74

Activations Density 0.855%

phrases related to advice or warnings

phrases indicating potential hazards or dangers

No Comments

No Known Activations