INDEX

Explanations

mentions of pilot programs or individuals who are pilots

oai_token-act-pair · gpt-3.5-turbo

instances of the word "pilot."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

TOP

-0.73

upon

-0.70

ormon

-0.70

rem

-0.69

Prev

-0.69

Newsletter

-0.68

 Debor

-0.67

 Vatican

-0.65

Else

-0.65

å¦

-0.61

POSITIVE LOGITS

 pilot

1.45

 Pilot

1.20

 pilots

1.09

selage

1.03

wings

1.02

ilot

1.02

isode

0.82

 cockpit

0.80

alysed

0.76

 hangar

0.76

Activations Density 0.007%

mentions of pilot programs or individuals who are pilots

instances of the word "pilot."

No Comments

No Known Activations