INDEX

Explanations

phrases related to sports news and events

oai_token-act-pair · gpt-3.5-turbo

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

.).

-0.95

.ï¿½

-0.95

.?

-0.92

."[

-0.90

.:

-0.83

().

-0.81

 goddamn

-0.80

.</

-0.79

$.

-0.78

.):

-0.77

POSITIVE LOGITS

 apologise

1.25

 favourites

1.19

 Â£

1.15

 apologised

1.14

 fortnight

1.14

 recognise

1.14

ITV

1.11

 whilst

1.11

 Newcastle

1.10

 Cardiff

1.08

Activations Density 0.816%

phrases related to sports news and events

No Comments

No Known Activations