INDEX

Explanations

words related to cordoning off or isolating areas

oai_token-act-pair · gpt-3.5-turbo

actions and states related to confinement or being cut off from others

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

alter

-0.67

 miracles

-0.64

 parity

-0.62

 Proced

-0.62

negie

-0.61

sonian

-0.60

 revolutions

-0.57

 Replacement

-0.57

 refurb

-0.57

 Sutherland

-0.56

POSITIVE LOGITS

 lest

1.03

gha

0.91

 airspace

0.86

lees

0.84

 altogether

0.83

 pores

0.81

camera

0.80

 entrances

0.79

jah

0.77

adow

0.75

Activations Density 0.476%

words related to cordoning off or isolating areas

actions and states related to confinement or being cut off from others

No Comments

No Known Activations