INDEX

Explanations

multi-part content

oai_token-act-pair · gpt-3.5-turbo

references to parts or episodes in a serialized format

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

ubs

-0.79

ampires

-0.78

gnu

-0.75

osponsors

-0.74

wolves

-0.74

Known

-0.73

ulkan

-0.72

orers

-0.72

ibia

-0.68

inges

-0.67

POSITIVE LOGITS

 puzzle

0.83

 documentary

0.77

 saga

0.77

 investigative

0.74

 sequential

0.74

 series

0.72

tenance

0.71

 investigation

0.70

 eviction

0.68

meal

0.68

Activations Density 0.063%

multi-part content

references to parts or episodes in a serialized format

No Comments

No Known Activations