INDEX

Explanations

references to rewards or incentives, often associated with tasks or targets

oai_token-act-pair · gpt-3.5-turbo

terms related to bounties and rewards for capturing or obtaining information about individuals

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

ppard

-0.89

 Reaction

-0.77

icken

-0.76

iber

-0.75

isive

-0.73

inking

-0.72

 Ange

-0.71

inem

-0.70

andem

-0.69

INK

-0.69

POSITIVE LOGITS

 bounty

1.44

 bount

1.28

 hunter

1.23

 hunters

1.04

 Bounty

1.03

 hunting

0.82

 reward

0.81

iless

0.78

hunter

0.77

 Hunters

0.76

Activations Density 0.019%

references to rewards or incentives, often associated with tasks or targets

terms related to bounties and rewards for capturing or obtaining information about individuals

No Comments

No Known Activations