INDEX

Explanations

concepts related to physical locations or points where movement or action begins or ends

oai_token-act-pair · gpt-3.5-turbo

references to various types of stops or halts in activities or processes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

venant

-0.76

ahime

-0.68

ale

-0.66

ilk

-0.64

WE

-0.59

otyp

-0.58

iw

-0.57

Lov

-0.57

iom

-0.57

central

-0.55

POSITIVE LOGITS

 stops

3.93

 stop

2.21

stop

1.88

 stopped

1.81

 stopping

1.80

 pauses

1.70

 Stop

1.64

Stop

1.55

 ceases

1.54

 STOP

1.45

Activations Density 0.007%

concepts related to physical locations or points where movement or action begins or ends

references to various types of stops or halts in activities or processes

No Comments

No Known Activations