INDEX

Explanations

words related to giving instructions or asking for information

oai_token-act-pair · gpt-3.5-turbo

phrases associated with authority and decision-making

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

isSpecialOrderable

-0.76

 guiActiveUnfocused

-0.72

NES

-0.68

bryce

-0.62

 hiatus

-0.60

quished

-0.58

ABC

-0.58

onyms

-0.58

 Emblem

-0.57

 fearing

-0.57

POSITIVE LOGITS

 behalf

0.85

 impartial

0.75

enance

0.71

 factual

0.70

ezvous

0.70

itatively

0.69

 agendas

0.66

udic

0.65

 superv

0.65

awi

0.65

Activations Density 0.643%

words related to giving instructions or asking for information

phrases associated with authority and decision-making

No Comments

No Known Activations