INDEX

Explanations

superlatives indicating extreme negativity such as "worst"

oai_token-act-pair · gpt-3.5-turbo

references to the term "worst," particularly in negative contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

arij

-0.79

alde

-0.75

ulton

-0.74

ijk

-0.73

iverpool

-0.72

dinand

-0.72

itialized

-0.71

mun

-0.70

ependence

-0.69

bara

-0.68

POSITIVE LOGITS

 nightmare

0.92

 imaginable

0.89

 worst

0.88

 Worst

0.86

 offender

0.86

worst

0.85

 offenders

0.79

EST

0.74

 nightmares

0.73

iary

0.73

Activations Density 0.011%

superlatives indicating extreme negativity such as "worst"

references to the term "worst," particularly in negative contexts

No Comments

No Known Activations