INDEX

Explanations

pattern matching for certain symbols or indicators like 'Ċ' and 'Note' within a document

oai_token-act-pair · gpt-3.5-turbo

notations or markers indicating player stats and other relevant sports data

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 eleph

-0.84

 citiz

-0.83

 ageing

-0.71

 proport

-0.70

 aspir

-0.70

 unnecess

-0.68

 hemor

-0.66

 compan

-0.66

 tremend

-0.66

 separat

-0.65

POSITIVE LOGITS

Coach

1.01

Players

0.98

Grade

0.96

ÂłÂłÂł

0.95

ÂłÂłÂłÂłÂłÂłÂłÂł

0.90

ESPN

0.89

Pred

0.89

Stats

0.89

Round

0.89

Wide

0.88

Activations Density 0.180%

pattern matching for certain symbols or indicators like 'Ċ' and 'Note' within a document

notations or markers indicating player stats and other relevant sports data

No Comments

No Known Activations