INDEX

Explanations

emotional and experiential terms

oai_token-act-pair · gpt-3.5-turbo

duplicate uses of the prefix "rel" that suggest relational or contextual connections

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

 Sail

-0.64

 Slate

-0.64

err

-0.64

Fax

-0.63

 ransom

-0.62

 Blow

-0.62

 sail

-0.59

 blasphemy

-0.58

 procedural

-0.57

 mileage

-0.56

POSITIVE LOGITS

aunch

1.43

aun

1.34

atable

1.33

apses

1.30

apsed

1.29

ented

1.25

ocating

1.23

ighting

1.18

ocations

1.17

aying

1.15

Activations Density 0.027%

emotional and experiential terms

duplicate uses of the prefix "rel" that suggest relational or contextual connections

No Comments

No Known Activations