INDEX

Explanations

instances where something is admitted or acknowledged

oai_token-act-pair · gpt-3.5-turbo

instances of the word "concede" and its variations

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

 shopping

-0.71

compan

-0.68

-0.65

 Influ

-0.64

 Lights

-0.64

shape

-0.64

 rein

-0.62

photos

-0.60

 Letters

-0.60

POSITIVE LOGITS

 concede

3.44

 conceded

3.41

 conced

3.12

 concedes

2.62

 concession

1.41

 acknowledged

1.33

 admits

1.30

 surrendered

1.26

 capit

1.22

 admitted

1.21

Activations Density 0.028%

instances where something is admitted or acknowledged

instances of the word "concede" and its variations

No Comments

No Known Activations