INDEX

Explanations

phrases instructing the user to input information

oai_token-act-pair · gpt-3.5-turbo

commands related to user input actions

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

externalActionCode

-0.74

bite

-0.74

 Nanto

-0.70

 Barrier

-0.69

ensation

-0.62

blem

-0.61

blow

-0.61

Reilly

-0.60

rise

-0.60

paren

-0.60

POSITIVE LOGITS

umeric

0.83

Password

0.78

abel

0.78

 passwords

0.76

otom

0.76

 captcha

0.74

ername

0.70

 phrases

0.70

 Submit

0.70

 manually

0.70

Activations Density 0.080%

phrases instructing the user to input information

commands related to user input actions

No Comments

No Known Activations