INDEX

Explanations

dates and locations mentioned in a specific format

oai_token-act-pair · gpt-3.5-turbo

various forms of punctuation and their associated contexts in textual data

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

lando

-0.83

 Schwartz

-0.80

zona

-0.76

 favor

-0.76

avorable

-0.74

ospons

-0.74

 favors

-0.73

ASHINGTON

-0.71

DOE

-0.71

 someday

-0.70

POSITIVE LOGITS

 apologise

1.12

Scotland

1.09

 apologised

1.05

 mould

1.01

 fibre

1.01

 centres

1.00

 realise

0.99

 recognise

0.99

 Premiership

0.98

 humour

0.98

Activations Density 0.291%

dates and locations mentioned in a specific format

various forms of punctuation and their associated contexts in textual data

No Comments

No Known Activations