INDEX

Explanations

terms related to covert activities or actions

oai_token-act-pair · gpt-3.5-turbo

terms related to secretive or undercover activities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

hello

-0.83

 Louie

-0.83

 Statue

-0.73

 Rating

-0.71

Assembly

-0.71

li

-0.71

tl

-0.70

Reference

-0.70

 Insurance

-0.69

Finish

-0.69

POSITIVE LOGITS

 covert

3.17

 clandestine

2.69

 illicit

2.00

 discreet

1.90

andestine

1.56

 stealth

1.55

 secretive

1.43

 candid

1.37

 nefarious

1.36

 paramilitary

1.35

Activations Density 0.044%

terms related to covert activities or actions

terms related to secretive or undercover activities

No Comments

No Known Activations