INDEX

Explanations

medical and disciplinary terms such as 'repro', 'admon', 'reprim', and 'rebuke'

oai_token-act-pair · gpt-3.5-turbo

terms related to reprimanding or criticizing behavior

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

arde

-0.86

emi

-0.73

ero

-0.73

forms

-0.69

omnia

-0.69

eper

-0.69

ieg

-0.68

ird

-0.68

eland

-0.68

isable

-0.68

POSITIVE LOGITS

 rebuke

1.04

 reprim

0.95

 admon

0.91

 repro

0.90

 questioning

0.88

dding

0.81

 menacing

0.80

>>\

0.79

æĢ

0.75

 chast

0.75

Activations Density 0.029%

medical and disciplinary terms such as 'repro', 'admon', 'reprim', and 'rebuke'

terms related to reprimanding or criticizing behavior

No Comments

No Known Activations