INDEX

Explanations

phrases related to having conversations or discussions

oai_token-act-pair · gpt-3.5-turbo

phrases related to discussions or conversations about political matters

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

:(

-0.92

)!

-0.83

!!

-0.82

Redditor

-0.79

??

-0.73

 Died

-0.73

 âĿ

-0.71

 Sadly

-0.71

 luckily

-0.71

 doesnt

-0.70

POSITIVE LOGITS

 vetting

1.31

 briefings

1.30

 memos

1.18

 vetted

1.14

 advisers

1.09

CIA

1.07

 Tillerson

1.05

 briefed

1.05

 confidential

1.04

 staffing

1.03

Activations Density 1.198%

phrases related to having conversations or discussions

phrases related to discussions or conversations about political matters

No Comments

No Known Activations