INDEX

Explanations

terms related to beliefs, assumptions, and acknowledgments

oai_token-act-pair · gpt-3.5-turbo

phrases that express common beliefs or widely held views

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

bound

-0.69

ĪĴ

-0.66

adra

-0.65

alos

-0.64

ubi

-0.62

Correction

-0.59

kay

-0.59

pan

-0.58

agraph

-0.58

manent

-0.58

POSITIVE LOGITS

isSpecialOrderable

0.84

 underestimate

0.77

 Occupations

0.75

 skept

0.74

Synopsis

0.71

 internationally

0.67

STD

0.65

 taboo

0.65

 amongst

0.63

 derog

0.63

Activations Density 0.172%

terms related to beliefs, assumptions, and acknowledgments

phrases that express common beliefs or widely held views

No Comments

No Known Activations