INDEX

Explanations

references to the name "Joe."

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

NESS

-0.93

hips

-0.87

 glim

-0.81

rawdownloadcloneembedreportprint

-0.75

mble

-0.73

arians

-0.72

ties

-0.71

iments

-0.71

ample

-0.71

igators

-0.68

POSITIVE LOGITS

 Biden

1.12

 Arpaio

1.06

 Russo

0.90

Rog

0.88

 Scarborough

0.86

 Lieberman

0.85

 Camel

0.84

pport

0.81

 Gibbs

0.79

 Frazier

0.79

Activations Density 0.027%

references to the name "Joe."

No Comments

No Known Activations