INDEX

Explanations

words related to expressing opinions or making requests

oai_token-act-pair · gpt-3.5-turbo

terms related to formal discourse and expressions of opinion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

orks

-0.71

 Saud

-0.64

iliated

-0.64

ÃŁ

-0.63

ecause

-0.62

ccording

-0.62

uum

-0.62

respect

-0.61

 airflow

-0.60

glomer

-0.59

POSITIVE LOGITS

fulness

0.85

 spree

0.78

 regarding

0.71

 exploits

0.70

 foray

0.69

 stance

0.69

iques

0.68

 rampage

0.66

lessness

0.65

 crusade

0.64

Activations Density 0.314%

words related to expressing opinions or making requests

terms related to formal discourse and expressions of opinion

No Comments

No Known Activations