INDEX

Explanations

the prefix "unf-" used to convey negation or reversal of an action

oai_token-act-pair · gpt-3.5-turbo

terms related to unfaithfulness and negative qualities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

anwhile

-0.83

hyde

-0.77

 Thornton

-0.72

uyomi

-0.72

ONES

-0.69

 Tackle

-0.68

 Notting

-0.68

OA

-0.67

Rin

-0.66

 Opera

-0.66

POSITIVE LOGITS

avour

1.14

athom

0.91

ortun

0.90

ortunately

0.89

requ

0.87

ishable

0.85

ruit

0.83

unf

0.81

riend

0.80

licted

0.80

Activations Density 0.004%

the prefix "unf-" used to convey negation or reversal of an action

terms related to unfaithfulness and negative qualities

No Comments

No Known Activations