INDEX

Explanations

words related to serious wrongdoing or intense negative actions

oai_token-act-pair · gpt-3.5-turbo

expressions of extreme wrongdoing or moral outrage

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ulton

-0.79

runner

-0.79

ym

-0.74

wrapper

-0.71

hner

-0.71

ather

-0.68

roe

-0.67

chrom

-0.66

runners

-0.66

upp

-0.65

POSITIVE LOGITS

 egregious

0.95

 abuses

0.90

 offender

0.89

 injustice

0.89

 heinous

0.88

 offenders

0.85

 violations

0.82

 viol

0.79

 injust

0.76

 egreg

0.73

Activations Density 0.009%

words related to serious wrongdoing or intense negative actions

expressions of extreme wrongdoing or moral outrage

No Comments

No Known Activations