INDEX

Explanations

words related to negative or insulting behavior

oai_token-act-pair · gpt-3.5-turbo

negative or derogatory terms related to individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

spring

-0.76

 Maintenance

-0.67

Source

-0.66

 Source

-0.63

 PowerPoint

-0.62

 Accuracy

-0.62

 Fathers

-0.61

 Annotations

-0.61

 Spiritual

-0.60

 Authority

-0.60

POSITIVE LOGITS

jer

1.26

 jerk

1.10

usalem

1.02

boa

1.01

ometer

0.88

¶ħ

0.86

nesday

0.85

icho

0.84

ometers

0.81

bucks

0.81

Activations Density 0.007%

words related to negative or insulting behavior

negative or derogatory terms related to individuals

No Comments

No Known Activations