INDEX

Explanations

words related to a specific publication and its daily edition

oai_token-act-pair · gpt-3.5-turbo

references to news articles and subscription prompts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 unaccount

-0.73

 elim

-0.68

nces

-0.65

VIDIA

-0.64

 outl

-0.63

ibal

-0.63

itsch

-0.63

emouth

-0.62

 extrap

-0.61

awaru

-0.59

POSITIVE LOGITS

 Newsletter

1.12

 newsletter

1.02

Magazine

0.85

letters

0.85

 Opinion

0.74

 Email

0.73

iPhone

0.71

ottest

0.70

email

0.67

 Weekly

0.66

Activations Density 0.058%

words related to a specific publication and its daily edition

references to news articles and subscription prompts

No Comments

No Known Activations