INDEX

Explanations

recommendations or instructions in a document

oai_token-act-pair · gpt-3.5-turbo

phrases indicating ideal scenarios or recommendations

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

 addictive

-0.71

itous

-0.70

AIDS

-0.70

 dare

-0.66

 injustice

-0.66

 menace

-0.65

 Traff

-0.65

abuse

-0.63

 threatens

-0.62

 wonders

-0.61

POSITIVE LOGITS

 preferred

0.84

 preferring

0.80

 Preferred

0.78

FontSize

0.73

uci

0.72

 prefer

0.72

 ideally

0.72

medium

0.72

 preferably

0.70

ittal

0.70

Activations Density 0.676%

recommendations or instructions in a document

phrases indicating ideal scenarios or recommendations

No Comments

No Known Activations