INDEX

Explanations

various forms of the word "exploit" and related terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

thood

-0.76

 mint

-0.75

ucket

-0.74

arat

-0.72

 basil

-0.63

kos

-0.63

hair

-0.62

prints

-0.61

 captcha

-0.61

 nodd

-0.60

POSITIVE LOGITS

 loopholes

0.92

 weaknesses

0.86

ulner

0.80

CVE

0.78

oise

0.78

oit

0.77

 exploited

0.77

ulnerability

0.71

 vulnerabilities

0.70

ously

0.70

Activations Density 0.113%

various forms of the word "exploit" and related terms

No Comments

No Known Activations