INDEX

Explanations

words related to spreading information, often with negative connotations such as propaganda or falsehoods

oai_token-act-pair · gpt-3.5-turbo

vocabulary related to the spread of information, particularly regarding conspiracy theories or misleading narratives

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

Negative Logits

ctic

-0.67

 citiz

-0.63

leans

-0.62

ney

-0.62

 Skydragon

-0.62

cut

-0.61

Rated

-0.61

*/(

-0.61

assian

-0.60

zbek

-0.59

POSITIVE LOGITS

 misinformation

0.95

 illusions

0.90

 havoc

0.86

 seeds

0.81

 disinformation

0.81

 falsehood

0.80

 propag

0.77

 propagate

0.76

 myths

0.74

 rumours

0.73

Activations Density 0.116%

words related to spreading information, often with negative connotations such as propaganda or falsehoods

vocabulary related to the spread of information, particularly regarding conspiracy theories or misleading narratives

No Known Activations