INDEX

Explanations

references to specific locations with the word "Hill"

oai_token-act-pair · gpt-3.5-turbo

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

 unres

-0.70

 Alonso

-0.69

 Lisbon

-0.68

 oppos

-0.66

Â·Â·

-0.64

————

-0.61

============

-0.60

âĢ¢âĢ¢âĢ¢âĢ¢

-0.60

 Bucc

-0.60

âĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪ

-0.58

POSITIVE LOGITS

iard

1.65

tops

1.25

top

1.24

side

1.19

bill

1.14

yer

1.07

arious

1.06

ibur

1.02

walking

0.98

marks

0.97

Activations Density 0.048%

references to specific locations with the word "Hill"

No Comments

No Known Activations