INDEX

Explanations

proper names - specifically the surname "Vasquez" in this case

oai_token-act-pair · gpt-3.5-turbo

names or mentions of individuals, particularly those with the surnames Vasquez and Raso

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

é¾

-0.90

puff

-0.90

rador

-0.81

eely

-0.74

ãĤ£

-0.73

Mexico

-0.70

iggle

-0.70

utton

-0.70

emouth

-0.69

 Objective

-0.69

POSITIVE LOGITS

Vas

2.14

Ras

2.09

Kas

1.99

Pas

1.98

pas

1.68

Sas

1.62

pas

1.58

Tas

1.56

kas

1.43

Yas

1.42

Activations Density 0.064%

proper names - specifically the surname "Vasquez" in this case

names or mentions of individuals, particularly those with the surnames Vasquez and Raso

No Comments

No Known Activations