INDEX

Explanations

comments or responses in online discussions

oai_token-act-pair · gpt-3.5-turbo

references to commenters and their opinions in discussions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 ÂŃ

-0.76

 federally

-0.63

 foundations

-0.61

 lucrative

-0.60

 aspir

-0.58

 engulfed

-0.58

 immersed

-0.57

 beneficiaries

-0.57

 domestically

-0.57

 nascent

-0.57

POSITIVE LOGITS

 commenter

1.26

 comments

1.15

 reddit

1.11

 replies

1.09

 comment

1.06

 commenters

1.03

 commented

1.00

 Quote

0.97

 reply

0.97

 thread

0.95

Activations Density 0.693%

comments or responses in online discussions

references to commenters and their opinions in discussions

No Comments

No Known Activations