INDEX

Explanations

negations or contradictions

oai_token-act-pair · gpt-3.5-turbo

negative phrases or negations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Advocate

-0.64

stakes

-0.63

 Statements

-0.61

 Liberties

-0.61

 Exposure

-0.61

 Surve

-0.60

 Zeal

-0.60

ç·

-0.59

kamp

-0.59

 Perspective

-0.59

POSITIVE LOGITS

icably

1.22

ifying

1.21

ching

1.21

ched

1.18

icing

1.15

epad

1.11

 realizing

1.10

 knowing

1.09

orious

1.02

 bothering

1.00

Activations Density 0.096%

negations or contradictions

negative phrases or negations

No Comments

No Known Activations