INDEX

Explanations

information or details that are considered exclusive or confidential

oai_token-act-pair · gpt-3.5-turbo

instances of the word "scoop" and related terms

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

audi

-0.69

 juvenile

-0.63

PRES

-0.63

 generals

-0.62

relations

-0.62

 aggrav

-0.62

 states

-0.62

 aggress

-0.61

TERN

-0.60

repeat

-0.59

POSITIVE LOGITS

 scoop

1.49

atform

0.88

sheet

0.83

oops

0.82

veland

0.82

hest

0.81

irtual

0.80

icular

0.78

icle

0.77

 Canaver

0.77

Activations Density 0.005%

information or details that are considered exclusive or confidential

instances of the word "scoop" and related terms

No Comments

No Known Activations