INDEX

Explanations

references to harmful or dangerous entities, such as killer whales or people

oai_token-act-pair · gpt-3.5-turbo

references to "killer whales."

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

ional

-0.91

ational

-0.83

ourced

-0.83

bles

-0.82

é¾

-0.80

yrinth

-0.80

ourcing

-0.78

edu

-0.78

OVER

-0.77

atile

-0.76

POSITIVE LOGITS

 killer

0.93

killer

0.92

 killers

0.84

 whales

0.82

 Killer

0.79

 spree

0.77

 instinct

0.75

 whale

0.73

knife

0.71

 Orange

0.71

Activations Density 0.028%

references to harmful or dangerous entities, such as killer whales or people

references to "killer whales."

No Comments

No Known Activations