INDEX

Explanations

terms related to deception or misinformation

oai_token-act-pair · gpt-3.5-turbo

references to misleading information or statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

mun

-0.80

area

-0.71

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

-0.71

aldo

-0.69

itar

-0.69

mega

-0.68

FM

-0.66

Merit

-0.66

dain

-0.66

ucha

-0.66

POSITIVE LOGITS

ingly

1.02

 misleading

0.92

 misled

0.87

 mislead

0.85

 deceive

0.81

 misrepresent

0.78

 statements

0.75

 tactics

0.74

 excuse

0.73

 disclosures

0.72

Activations Density 0.015%

terms related to deception or misinformation

references to misleading information or statements

No Comments

No Known Activations