INDEX

Explanations

phrases related to interruptions or emphasis in a text

oai_token-act-pair · gpt-3.5-turbo

sections or elements that are separated by a delimiter

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 behavioural

-0.95

 colourful

-0.92

 ÂŃ

-0.85

 favourite

-0.85

 aluminium

-0.84

 analys

-0.84

 colour

-0.81

 analyse

-0.81

 humour

-0.80

 realise

-0.80

POSITIVE LOGITS

--

3.73

)--

2.68

---

2.50

!--

2.29

.--

2.27

----

2.20

--

2.07

-----

1.89

``

1.80

-->

1.78

Activations Density 0.010%

phrases related to interruptions or emphasis in a text

sections or elements that are separated by a delimiter

No Comments

No Known Activations