INDEX

Explanations

references to human-related concepts or attributes

oai_token-act-pair · gpt-3.5-turbo

references to human beings and their behaviors

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

arella

-0.86

abb

-0.73

anwhile

-0.71

REP

-0.71

Ã¨re

-0.70

rypt

-0.70

 ..........

-0.70

forth

-0.69

illary

-0.69

kick

-0.69

POSITIVE LOGITS

 beings

1.25

 readable

1.03

oids

1.02

 embryonic

0.91

 civilization

0.88

itar

0.82

oid

0.82

 fingert

0.82

zee

0.81

itarian

0.81

Activations Density 0.024%

references to human-related concepts or attributes

references to human beings and their behaviors

No Comments

No Known Activations