INDEX

Explanations

proper nouns and locations mentioned in news or articles

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

 reservations

-0.75

 unrecogn

-0.73

 unmatched

-0.69

 deciding

-0.68

 debilitating

-0.66

isSpecialOrderable

-0.65

 stricken

-0.64

 advis

-0.64

 backward

-0.64

 conflicted

-0.64

POSITIVE LOGITS

 Shutterstock

1.25

photos

1.19

 Photography

1.12

 Archives

1.04

ulton

1.00

Images

0.99

utterstock

0.99

 Courtesy

0.97

 IMAGES

0.97

 Images

0.96

Activations Density 0.205%

proper nouns and locations mentioned in news or articles

No Comments

No Known Activations