INDEX

Explanations

percentages or numerical increases/decreases

oai_token-act-pair · gpt-3.5-turbo

phrases indicating statistical increases or decreases

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Founding

-0.58

 Centauri

-0.56

 Fenrir

-0.56

 severed

-0.56

 Andromeda

-0.55

mop

-0.54

 Establishment

-0.54

å£«

-0.53

çĶŁ

-0.53

 dinners

-0.53

POSITIVE LOGITS

graded

1.04

 marginally

1.01

 sharply

0.99

grading

0.96

grades

0.89

 slightly

0.84

 significantly

0.83

0.81

 considerably

0.80

 markedly

0.80

Activations Density 0.050%

percentages or numerical increases/decreases

phrases indicating statistical increases or decreases

No Comments

No Known Activations