INDEX

Explanations

highly specific locations or objects related to news events

oai_token-act-pair · gpt-3.5-turbo

phrases indicating discovery or investigation of substances or materials

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

gencies

-0.86

 thereof

-0.83

Strength

-0.78

catentry

-0.76

relations

-0.74

DragonMagazine

-0.74

relation

-0.74

ny

-0.72

Anyone

-0.71

Definition

-0.71

POSITIVE LOGITS

 hundreds

1.04

 dozens

1.03

 thousands

0.98

 radioactive

0.89

 aborted

0.89

 discarded

0.88

 endangered

0.88

 giant

0.87

 genetically

0.86

 contaminated

0.86

Activations Density 0.431%

highly specific locations or objects related to news events

phrases indicating discovery or investigation of substances or materials

No Comments

No Known Activations