INDEX

Explanations

phrases related to searching for trustworthy news

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a search for reliable information or news

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

âĹ¼

-0.75

Lago

-0.75

anas

-0.74

vasive

-0.71

wu

-0.71

upuncture

-0.71

arser

-0.70

SPONSORED

-0.69

unes

-0.69

scribed

-0.68

POSITIVE LOGITS

 suspic

0.87

 adolesc

0.73

allery

0.69

Looking

0.68

 Opportun

0.68

finder

0.68

 Glass

0.68

 cursor

0.66

 pregn

0.66

 citiz

0.65

Activations Density 0.030%

phrases related to searching for trustworthy news

phrases indicating a search for reliable information or news

No Comments

No Known Activations