INDEX

Explanations

references to technical details and specific visual descriptions, possibly related to scientific or aerospace contexts

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

images and visual representations in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_post

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_post

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

 doctrines

-0.70

 quotas

-0.68

 curfew

-0.66

 affairs

-0.60

 attrition

-0.60

 therapies

-0.59

 disband

-0.59

 administrations

-0.59

 quota

-0.58

ivism

-0.58

POSITIVE LOGITS

image

1.14

 Image

1.11

 Thumbnails

1.11

 Notice

1.07

Picture

1.04

Image

1.04

picture

1.04

Credit

1.04

photo

1.03

Screenshot

1.02

Activations Density 0.600%

references to technical details and specific visual descriptions, possibly related to scientific or aerospace contexts

images and visual representations in the text

No Comments

No Known Activations