INDEX

Explanations

references to participation in activities or events

oai_token-act-pair · gpt-3.5-turbo

instances of the word "played."

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

 specification

-0.67

 quarterly

-0.59

 mail

-0.58

ordinate

-0.57

IPS

-0.57

aer

-0.56

 institute

-0.56

 commission

-0.55

 deposit

-0.55

 lash

-0.54

POSITIVE LOGITS

 played

3.62

 Played

2.76

played

2.43

 plays

2.02

 play

1.97

 playing

1.86

Play

1.80

 coached

1.72

 Playing

1.70

Playing

1.70

Activations Density 0.020%

references to participation in activities or events

instances of the word "played."

No Comments

No Known Activations