INDEX

Explanations

words related to television or broadcasting

oai_token-act-pair · gpt-3.5-turbo

references to television

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 brute

-0.75

 plank

-0.67

 sage

-0.64

 Hague

-0.64

 sorry

-0.64

gra

-0.63

 gener

-0.62

 pers

-0.60

 ease

-0.60

paw

-0.59

POSITIVE LOGITS

TV

4.20

tv

2.52

TV

2.03

levision

1.90

tv

1.88

 Television

1.71

 television

1.66

CTV

1.59

TVs

1.54

FM

1.43

Activations Density 0.017%

words related to television or broadcasting

references to television

No Comments

No Known Activations