INDEX

Explanations

statements related to legal or investigative processes

oai_token-act-pair · gpt-3.5-turbo

phrases related to legal processes and statuses

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 hates

-0.66

 regrets

-0.62

 feels

-0.61

 regretted

-0.60

 detrim

-0.60

 spends

-0.59

 Stats

-0.58

 apologizing

-0.58

 retains

-0.57

 ratings

-0.57

POSITIVE LOGITS

 arisen

1.37

 arise

1.34

 ensued

1.32

 emerge

1.30

 exist

1.23

 emerged

1.18

 occur

1.13

 accompany

1.03

 existed

1.02

 occurred

1.02

Activations Density 0.255%

statements related to legal or investigative processes

phrases related to legal processes and statuses

No Comments

No Known Activations