INDEX

Explanations

references to exceptions in rules, systems, or situations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "exception."

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

lab

-0.69

ching

-0.66

uld

-0.65

yang

-0.63

roc

-0.63

roph

-0.62

odium

-0.62

Lab

-0.61

cong

-0.61

itance

-0.60

POSITIVE LOGITS

 exception

1.09

 exceptions

1.06

except

0.85

witz

0.73

 Exception

0.73

ishly

0.72

ptions

0.71

DERR

0.70

 objections

0.70

flake

0.69

Activations Density 0.013%

references to exceptions in rules, systems, or situations

instances of the word "exception."

No Comments

No Known Activations