INDEX

Explanations

phrases related to negative actions or intentions, such as undermining, discrediting, or destroying

oai_token-act-pair · gpt-3.5-turbo

references to actions that weaken or damage authority or credibility

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Reflex

-0.63

lov

-0.63

area

-0.63

ëĭ

-0.62

enne

-0.61

cise

-0.61

ann

-0.60

atar

-0.60

andro

-0.60

onna

-0.59

POSITIVE LOGITS

 havoc

0.91

 undermin

0.86

 undermining

0.78

 undermine

0.73

ments

0.68

ionage

0.65

 undermines

0.65

amental

0.64

lessly

0.63

 guiActiveUn

0.63

Activations Density 0.044%

phrases related to negative actions or intentions, such as undermining, discrediting, or destroying

references to actions that weaken or damage authority or credibility

No Comments

No Known Activations