INDEX

Explanations

words related to accusations of wrongdoing and misconduct

oai_token-act-pair · gpt-3.5-turbo

references to misconduct or unethical behavior

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

population

-0.78

etically

-0.75

etic

-0.72

pop

-0.69

etics

-0.68

orph

-0.66

birth

-0.65

ixed

-0.64

icultural

-0.63

Wel

-0.62

POSITIVE LOGITS

 wrongdoing

1.27

 misconduct

0.97

 unfocusedRange

0.90

 perpetrated

0.88

 mishand

0.86

 allegations

0.86

 disclosures

0.85

 Disclosure

0.85

 involving

0.81

 disclosure

0.81

Activations Density 0.020%

words related to accusations of wrongdoing and misconduct

references to misconduct or unethical behavior

No Comments

No Known Activations