INDEX

Explanations

words related to sensitive or confidential information

oai_token-act-pair · gpt-3.5-turbo

terms related to secrecy and serious incidents

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Ì¶

-0.88

ecause

-0.84

ocrates

-0.81

olkien

-0.77

Ô

-0.74

[_

-0.73

cius

-0.71

akespeare

-0.71

olean

-0.70

ventus

-0.68

POSITIVE LOGITS

 ordeal

1.14

 incident

1.11

 arrangement

1.06

 portion

1.06

 event

1.03

 endeavor

1.03

 affair

1.02

 probe

1.02

 operation

1.01

 edition

1.00

Activations Density 0.419%

words related to sensitive or confidential information

terms related to secrecy and serious incidents

No Comments

No Known Activations