INDEX

Explanations

words related to hexes, charms, or magic spells

oai_token-act-pair · gpt-3.5-turbo

references to curses or negative supernatural concepts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

å§«

-0.93

olitan

-0.79

nington

-0.79

anooga

-0.77

endment

-0.75

visors

-0.75

ãĥ´ãĤ¡

-0.74

ighters

-0.73

regor

-0.73

axies

-0.73

POSITIVE LOGITS

 curse

0.99

 curses

0.91

 cursing

0.82

 cursed

0.80

bones

0.74

hammer

0.74

 fate

0.70

 wrench

0.69

vine

0.68

crow

0.68

Activations Density 0.012%

words related to hexes, charms, or magic spells

references to curses or negative supernatural concepts

No Comments

No Known Activations