INDEX

Explanations

the name "Top eka" in repeated contexts

oai_token-act-pair · gpt-3.5-turbo

references to specific locations and names related to events or people

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

äº

-0.74

 Hannibal

-0.65

Raj

-0.63

 sequest

-0.62

 sanity

-0.62

Hep

-0.61

 Cher

-0.61

 pres

-0.61

RAM

-0.60

 Jere

-0.60

POSITIVE LOGITS

eka

1.45

OE

1.36

 McGr

1.34

 Canterbury

1.20

 McCabe

1.09

 Morales

1.04

 proliferation

0.97

oco

0.97

opath

0.96

 Bauer

0.95

Activations Density 0.054%

the name "Top eka" in repeated contexts

references to specific locations and names related to events or people

No Comments

No Known Activations