INDEX

Explanations

terms related to harmful or destructive actions or intentions

oai_token-act-pair · gpt-3.5-turbo

terms related to malicious behavior or intentions

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

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

 Transition

-0.80

bach

-0.74

omics

-0.73

 plateau

-0.71

 basics

-0.70

 ratios

-0.69

 pickups

-0.68

inav

-0.66

undrum

-0.66

Reviewed

-0.63

POSITIVE LOGITS

 malicious

3.47

 malice

2.23

 nefarious

1.90

 sinister

1.54

 hateful

1.52

 dishonest

1.49

 fraudulent

1.43

 mischief

1.42

 vicious

1.41

 lewd

1.37

Activations Density 0.036%

terms related to harmful or destructive actions or intentions

terms related to malicious behavior or intentions

No Comments

No Known Activations